Click for next page ( 150


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 149
APPENDIX E Review of Obstetrical Risk Assessment Methods Beatrice/. Selwyn Thirty-three articles on obstetrical risk assessment methods reported in the literature are described in Table 1. The studies represent the uae of 19 different scoring systems. Nine studies were published prior to 1973J the reaaining 24 were reported in 1973 or later. A historical review of the risk assessment literature by Robel (1976) included 7 of the systems in Table 1. In the pre-1973 era, 3 of the most widely used risk assessment methods were reporteds Apgar (1953), Goodwin et al. (1969), and Nesbitt and Aubry (1969). In 1973, Robel's method was first published (Robel et al., 1973) and now joins the others as one of the most widely used methods. ATTRIBUTES OP THE METHODS The methodology used in the studies described in Table 1 varies from saaplea of highly selected groups of women (James et al., 1976J Kaminski et al., 1973) to collections of large numbers of consecutively sampled WOBen (Coopland, et al., 1977J Robel, et al. 1973J Nesbitt and Aubry, 1969). Methodology influences findings because the occurrence of an outcome in the study group is affected by the characteristics of the group chosen for study. The predictive accuracy of the system may also be affected by the generalizability of findings from one group to another. Most of the methods require observations of the woman being scored, and a few require interviews with the woman (Robel et al., 1973J laainski et al., 1973). The number of characteristics ascertained varies from method to method: 155 were assessed by Effer (1969)J 126 by Robel et al. (1973)J 123 by Stembera et al. (1975)J approximately 45 by Nesbitt and Aubry (1969)J and approximately 21 by Goodwin et al. (1969). The categories of factors assessed in all methods are similar. They include demographic data, socioeconomic data, data based on past preg- nancies, medical history, present pregnancy and, in the more recent studies, data on fetal heart rate and uterine contractions from electronic monitoring. All authors have baaed their selection of factors on the existing information concerning variables associated 149

OCR for page 149
TABLE 1 Attributes of Obstetrical Risk Assessment Methods Study Matboda Risk Aa ..a...nt Matbod llo. llo. stye When Collected ScorinG of of Data Sub- rec- Pre- Intr ... Manner of Outco.e of . .fa ranee Dedgn Source a jecta Title tor a Pectora Included natal partua Neonatal Scoring Interest AP9ar, U53 Coneec:utive Obeervation! 1,760 AP9ar 5 Heart rate, reapiratory 1 minute Points Acidoaia1 birtha, infant acore effort, reflex. irrita- after birth giveno 110rtality delivery bility, auacla tone, 0. bad color 2 • 9-d aua points, 10 1a beat, arbitrary Prachtel, Sa.ple 30' lladical 1,378 Obatat- 42 Obatetrical and IIOCio- Yea Yea 3 and 10 1 point for Abnor.. l U67 higb-riak vo.en, record~ ~ deal ec:onoaic (SBSI variables, daya after each non- neurology Ul study groupo Score 58 neurological signa birth optlaal 0 infanta born factor, in boapital aua points Larks and Coneec:utive Obaarvatioa 2,028 54 DeiiOgraphic characterie- Yes Multivariate AP9ar Larks, lt68 births tics, blood presaure of acora acore1 110thar, bioelectrical 110rtallty -••urea »asbitt and Coneec:utive Obaarvatlon 1,001 MCB care !,45 DeiiOgraphic character- Only Arbitrary ._r pari- Aubry, 1969 ward patients, Indesi istics parity, obatatri- veightso natal out- care giYSn , ... i- cal history, aedical 0 • good ~e without knowing objactiYS history, IIOCioeconollic 30 • bad! acore greding status, aaotional sua points syat•l status and subtract fraa 100 Wilson and RandOII 118J111le Obaarvation 148 MCB care 21 s... Only saae Sue Sill, 1973 of ~king~, and recorda Indax of aa.ple of 150 »aabitt dellveriaaJ and Aubry, coabinad the 1969 t110 aa.ple a Aubry and Coneec:uti,. Obaervation 450 MCB ears 21 s - . with tba Yea Yea s - . but s- PaMington, adllisaioas to and records Index of addition of .. tunal, with Labor U73 labor and Nallbitt fatal, and placental Index, aua, delivery, and Aubry, factor a and subtract alxed clinic U6!1 plus 5 fraa 200 and privata Labor Index patients

OCR for page 149
Goodwin et al., Several study ObHrvation 1136 Antepartua !,21 laae1ine and obsta- Yea Arbitl:ary Perinatal uu groups are and records retal Risk trical history, and based 110rtality co.blned 1 peri- Score preMnt praguncy, on litera- natal deaths, gestational age ture, weight at birth factors ward dellverles, 0-10 (0 • and private Bebb et al., 11180 CaMs All preutal patients of Observation 17,270 Antepartua retal Risk !21 a... , except no uH of gestational age Yes good), sua Saas .... various MDs Score of in region Goodwin, et al., ·- 11169 Coopland et al., All wc.en Records 5,U9 Modified !21 a... , except aoae Yes Weights are U77 adaitted to Goodwin, deaagraphic factors 0-3 labor and Tlloaes Yeh et al., 1977 w-en aonitored during labor Records 266 1969 Goodwin et alo 1 19611 !21 a... ,plus fetal heart rate and uterine con- Yes Yes Weights are 0-4 .... plus fetal traction ·- 110nitoring Morrison and Olsen, 111711 Deliveries in 80 hospitals ObMrvation 16,733 Goodwin et al., 19611 !21 ·- Yes On ad- aission in labor Weights are 0-3 Butler and Alber•n, All births in ObHrvation l week in Great recorda, 17,024 Rona 6 Age, parity, Yes Yea Multivariate score for .... 1969 Britainr still- birth and births and death car- 7,851 social class, height, each factor and sua .... Ul neoutal deaths tificates pre-eclaapsia, and for 3 110nths 8110king Alberasn and Goldstein, 1970 rr- Perinatal Mortality Surveyr iden• tify children at 7 years ·- 12,083 Hone- 167 with handicaps 3 Parity 4+, adverse aethod of delivery, neonatal illness in 1st week of life Yes Yes Yea One point for preHnce of factor Handicap with and without handicaps Bffer, 1969 All adaissions Records 211 Prognostic 155 Pathology, Assigned Not clear, Apgar to high-risk 350 Risk Score test results, at labor aasuae 1 pregnancy unitr deaagraphic and au, point if rancSo. naple past and preHnt preHnt of thoH with obstetrical probl... standard care

OCR for page 149
Table 1' Continued Study Methode Risk As . .s . . .nt Method ao. ao. Stage When Collected Scoring of of Data Sub- rae- Pre- Intra- Manner of Outca.e of aeference Design Sources jecte 'l'itle tore ractora Included natal partua Neonatal Scoring Interest Rantakallio, All births in Recorda and 11,391 Hone U! Biological factora, Yes Diacr i11inant Poor neo- 11169 one year intervie'M!! in factors of aother, function, natal out- lat .acioeconoaic status Eigenvector c-e anal- weights given yaia1 each factor 27 in l final RObel et al., All v-.n c..- Intervi8V and 725 Screening 51 Prenatal, intrapartua, Yea Yea Yea Arbitrary Perinatal 1973 ing to clinic1 observation to predict 40 and neonatal weights of 110rbidity acrHned, but high riak 35 1 (good), and attending MD did not know neonate a 5 and 101 aortality .... Ul acore pre- acore N natal aa . .an, acore intra- partua as eu11 ·- of pointe Bobel, 1976 s... , update Sue 1,417 llobel'a 35 Yea Yes Yea Saae Sue aodified High-riak n - a... 60 Robel'• 35 Neonatal factors Yea Yea Yea s ... Morbidity nate . . tched 52 aodified in lat 2 with l~riak years of RObel, 11179 Sokol at al., neonate s . . . aaaple Conaecutive .... Intervi.., and 1,417 Sobel'• aodified 1,275 Sobel'• 21 18 49 Prenatal, intrapartu11 Prenatal, Yea Yes Multivariate acorea given life s... Yea Yea s ... s ... 1977 Sokol et al., 1979 deli varia a Delivery with perinatal deathJ delivery with- observation s- aodified 143 Robal'a- aodified 36 36 intrapartu• s... Yea Yes Yea .... sa- out perinatal death 5,235 Chik at al., Conaecutively Intervi.., and 4,500 Sobel'• 36 s ... , but with Yea Yes Diacri11inant s... l97t delivered observation aodified 17 fetal aonitoring function - n with factors acorea fetal aoni- toringr He Sokol at al., 1979 L

OCR for page 149
Wlntel'• et al •• 1.171 8l.aok and Hi- panic - n %n~ei'Y1ew record• and 62 llobel'• ·- ·- Yee Yee Yea ··- .... bavlnv Uve infanta Ilea•-~ Ra OCR for page 149
Table 1 Continued Study Methoda Risk Aaaea... nt Method No, No. Stage When Collecte4 Scoring of of Data Sub- rae- Pre- Intra- Manner of Outc011e of Reference Design Source a jecta Title tors Pactora Inclu4e4 natal partu11 Neonatal Scoring Interest Cora4ello et al., Unknown1 None 41 Not apeciUe4 Yea Yea Not Mortality 1975 retroapecti ve Unknown 1,067 apecitie4 prospective Unknown 230 Ala anCI Neurologically ObHrvation Braselton, 1975 auapect infanta 53 Braselton 46 Behavioral, Yea Behavioral Neonatal followe4 for Bxa• neurological, factors 9et JDOrbi4ity 7 yean vi9or anCI attention, 9 points, .otor activity, an4 reflex tone autono.ic ..aaurea 9et reaponae 4 pointa1 BUll Ja..a et al., w-en elec- Recor4a 665 PHR-UP9. 10 Characteriatica of Yea Use4 Neonatal 1976 tronically Monitoring fetal heart rate an4 Receiver morbi4ity .onitore4 Score uterine preaaure Operator .... Ul Character- istic Curve anCI clinical • experience to vei9ht factors l'e4rick, 1976 CaHa • apon- Unknown 283 None 10 Dnoographic, Yea Calculate Preutur- taneoua pre- previoua pregnancy relative ity ter• birtha history risk for each factor~ eo.peen auple 510 •ultiply of singleton relative birthaa retro- risks for apective score Uwar4a et al,, WC..n 4eliver- Obaervation 2,085 None 67 Dnoographic, Yea Yea ArbitraryJ Neonatal 1!17!1 i ng conaec~t- obatetric, baH4 on .orbi4ity tively in ~ical, anCI i!lpOrtance anCI hoapital other to outc,.., .ortality wights • 1 (9oodl to 7 Jl)baervati- • 4ata collecte4 by ex-ination of the .otber of the infant, iaecor4a • 4ata collecte4 throuvh the uH of exiating recor4aa ~ical, birth certificatea, 4eath certificatea, etc. £Maternal Chil4 Bealth Care In4ex • MCB Care In4ex. !Interview • the riak inatru.ent require• 4ata obtained 4irectly fro. the other .othar•-.ora than what ia routinely collecte4 in hoapitals. SAl in firat analyaia, 27 in final analyaia, 1BPLII • h iatory, prenatal, labor, neonatal. lrsa-ur • fatal heart rata, uterine praaaure.

OCR for page 149
155 with outcomes such as perinatal mortality or morbidity, rather than on data concerning maternal mortality or morbidity. The stage of pregnancy when factors are assessed can affect the accuracy of the prediction and the timeliness of intervention or prevention (Chng et al., 1980). Of the 19 systems, 15 require assess- ment of the woman prenatallyr 8 require assessment of the woman only during the prenatal period, e.g., Nesbitt et al. (1969), Goodwin et al. (1969). Thirteen methods include intrapartum factors, and six assess the neonate. Two scoring methods are used solely for assessment of infants: the Apgar Score (Apgar, 1953, 1966) and the Brazelton Method (Brazelton, 1973). Robel (1977, 1978), Robel et al. (1973, 1979), Prechtl (1967), and Stembera (1975) provide for collection of informa- tion during the entire pregnancy, i.e., prenatally through labor and delivery, including assessment of the neonate's characteristics. The manner of scoring each risk factor is handled uniformly by arbitrarily assigning a weight based on reports in the literature and on experience. Most of the scoring systems are easy to user indeed, this characteristic is frequently cited as a criterion in developing the system. Even in systeas in which scores were obtained by using a •ultivariate technique, arbitrary decisions were made at some point, e.g., level of seriousness of risk (James et al., 1976J Stembera et al., 1975). In some of the methods, attempts were made to accommodate weights to different outcome factors by using multiple regression techniques and discriminant function analysis (Donahue and Wan, 1973r Robel, 1979r James et al., 1976r Rantakallio, 1969). Biochemical measures and fetal monitoring are occasionally used together with existing risk assessment methods. Yeh et al., (1977) coabined them with the suggestion of Goodwin et al. (1969)J Chik et al. (1979) did the same with the Robel (1973) method. James et al. (1976) developed a scoring system for fetal monitoring and uterine contraction data. Appropriate use of biochemical indices to predict outcome re- quires detailed knowledge of the biochemical process (Tulchinsky, 1980). A thorough understanding of findings concerning fetal heart rate and judicious decisions about their clinical importance are necessary for successful prediction of outcomes (Chik et al., 1979). PREDICTIVE POWER OF THE MB'l'HODS Evaluation of the ability of a screening test to predict an outcome successfully entails knowledge of the test's sensitivity, specificity, and predictive value as well as information on the frequency of occur- rence of the outcome (Table 2). Many of the reports referenced in Table 2 did not contain the requisite information. Therefore, the screening parameters presented in the table were calculated from the data presented by the authors. In some instances numbers were extracted from graphs in which only percentages were displayed. An attempt vas aade to standardize all parameters for purposes of comparison. Lesinski (1975) has reviewed the risk assessment literature emphasizing the clinical elements in screening. The following comments are focused on

OCR for page 149
TABLE 2 Predictive Power of Obstetrical Risk Assessment Methods Ability of High- and Low-Riak Scoraa to Predict a Given Outca.e~ Prevalence of --- Riak Neonatal C~licationa Low Birth weisht ILBMI Perinatal Death Titlea and 'riak low- ' low- ' low- riak risk Reference Scorea for Rlak (t) ,,, high Mdiu. 1- (t) ,,, inci aena (t) ,,, ,,, ,,, apec fal+ fal- with probl- ,,, ,,, ,,, inci aena a pee fal+ (t) ,,, fal- with ,,, ,,, ,,, ,,, inci sena spec fal+ fal- with proble• l'l proble• Apgar, U53 ApcJar 6 l8 76 Score a 1.2 92 77 23 8 0.1 high • Q-2 (Neonatal only) Md • 3-7 ~ • 8-10 Prechtel, Obatetri- 12 &8 19 38 16 23 10 14 28 1967 cal ecorea (Neurological only) high • 7+ . .d • 2-6 1- • 0-1 .... U'l Larka and No title• Rot known 0\ Larka, Multi- 1968 a 2 for stillbirth • 41~ variate a 2 for death in 2 years • 41' 8COrlng ...bitt and MCB Care 30 3t 31 6 47 32 28 19 4 ll 43 33 27 23 10 3 43 31 29 27 2.6 Aubry' 196!1 Index a high • 0-70 Md • 71-84 1- • 85-100 Wlleon and MCB Care 8 72 20 Bill, U73 Index a 17 6 22 8 15 13 high • 0-40 Md • 41-89 ~ • tO+ Aubry and MCB Care 21 7t Pennington, 8.4 63 82 l7 37 4 Index plua U73 Labor Inclexa high • none 1- • none Goodwin, Antepart~a 7 6!1 24 et al., Petal Riek 15 46 28 0 0 0 U6!1 Score• high • 7-10 Md • l-6 1- • 0 Bebb et al., Antepart~a 10 75 15 ltiO Petal Riak 2.6 86 15 8 l 0.2 Score a high • 4+ Md • l-3 1- • 0 L

OCR for page 149
~1and Ante..-rt.u. •ot. known • hi9heat cl•k wlt.h lt. - so• ' hl9heat rl•k with it .• 1U t hi9he•t ri•k wlth it • 11.0' et; al., Petal lliall ' 1oweat riak with it • '' ' 1oweat riak with it • lt t 1oweat riak with it • 0.4t lt77 Score• IIOdified higheet• 7+ high • 3-6 low • 1-2 loweat • o Yeh et el., Antepertua 28 72 t high riak with it • 22t 1977 l'etel Rial! t low riak with it • 13t Score, of fetal - i - toring• high • 4+ low • G-3 Mor:riaon Ante~rtua 19 81 1.9 70 82 18 30 0.7 end Ol .. n l'etd Rial! lt79 Score• high • 3+ low • 0-2 Butler end No title• Not known Bigheat riak • 85/1,000 deeth Albe~n, Multiveriete rete 1969 Score of 800 Loveat riak • 8.9/1,000 deeth (high risk) rete to -800 (low riakl Albe~n No title• 13 87 1.4 26 87 13 74 1.2 endGold- high • (Bandicepa only) etein, 1970 preHnce of eny 1 factor ~ Bffer, 1969 Prognoatic (high riak group) 38 Ul Risk Score• 46 54 55 62 38 45 31 (1 •inute Apger only, high-riak ..... high • 51+ IRa'*- group) group only) low • G-50 11 89 Rentelldlio, 14 86 2.4 30 87 13 70 2 1969 Diacri•inent Function Score• high • prob- ebility of poor out- COIH ia 50t Robel Scr ..ning 16 20 18 46 16 36 51 12 19 6 4 59 48 14 4 0.3 et el., 1973 to Predict BB LB HL LL Bigh-Riak lleonet••• high • 10+ lov • G-9 4 group• high/high with pre- low/high netel end high/low intre~rtua lov/low .:ore a

OCR for page 149
TABLE 2 Continued Ability of High- and Low-Riak Score• to Predict a Given Outcome! Pravalance of --- Rial! •-natal C!!!!l!licationa Low Birth wei9ht !LBW! Per ina tal Death • low- t low- t low- Titl.. and riak riak riak Reference Soor•• for Riak ,,, ,,, ,,, high MdiUII low inci ••n• •pee (t) fal+ (t) fal- with probl- ,., "' ,,, ,,, ,., inci aena apec fal+ fal- with problea (t) ,,, ,., ,,, ,,, inci aena apec fal+ fal- with problea "' "' "' llobe 1, 1!176 llobel'• 16 23 45 16 16 37 51 12 18 6 3 54 47 14 2 0.2 updated HH LBBL LL High Low sue (Group• defined by Poor weight gain 48 7 being at high or low riakl Robel, 1!179 IIObel' 81 18 82 54 72 28 46 12 ~ high • 10+ (Prenatal factor• only) U'l C» low • 0-9 94 28 72 6 6 (Prenatal and intrapartUII) Sokol Robel'•• 26 20 23 31 3 82 31 24 0 0 et al., high • 10+ 1977 low • 0-9 4 group• with pre- natal and intrapa~:- tua ~:iaka Sokol Robel'• Ca..-ca.pad110n et al., aodifiedl U79 (not known) Bigh-~:iak group • 46t death rate Low-dak group • 12t death rate Chik llobel'•· Not known 33 93 87 13 7 4 et al., aodifiedl (One ainute Apga~: a• outcoae1 1979 diacdai- ~:iak end aonito~:ing data) nant func- tion aco~:ea UHd Winter• llobel'•• 41 59 48 52 70 30 48 39 et al., high • 40+ 1979 low • 0-39 lteaane~: social- 55 45 u 72 48 52 28 7 2.2 75 46 54 25 1 et al., Medical Rial! (Infant death• only) 1973 Boor•• high • 2+ handicapa, l aocial • l Mdical L

OCR for page 149
•-lnakl No ~1~~-· 16 22 62 39 63 3. ·~ 4 2 et. al., •ul&ivaci- e.5+ o-5 lt73 •t.•80oc•• 0.5 hl9h ecoce (Scoce foe low • hl9h rlek1 binh -19ht aiiOUnt ae outcc.e) variee with ou~e Donahue and 'l'Otal Riak 25 50 25 33 43 29 16 16 22 lfen, 1973 Score, (Neonatal deeth• only) •ultivuiete ecore quer- tilee UMd for rieka high • upper 251 low • lower 25\ Haeri No titlea 12 88 3 28 ag 11 72 2 et al., 3 eyetHe, 1974 ueed herea high • 4+ low • 0.3 Helliday Haeri'e 26 36 38 7 64 40 23 13 2.5 1.4 67 38 25 6 0.2 et al., Scoring (Neonatal tranefer to lerge hoepitel) (Neonatel deathe only) 1980 Syete•• high • 7-16 ...s • 4-6 low • D-3 Stellbera et al., HPLH£ &corea llot known Popule- Morbid- Popule- Perina tel .... Ul 1975 high • HP 30+ tion (I) ity (I) tion I + infent HPLII 40+ •ultivariate H+P H+P+I:t+N 16 u 31 51 H+P 16 deathe • 40 "' ecoring H+P+I:t+N u 73 Coradello No Titlll llot known? 19 58 86 l4 42 10 et al., high • 51+ (Riek aeMeeed at delivery) 1!175 low • o-5o 6 77 78 22 23 2 (Riek aeMeeed prenatally) Ale end Bra•elton 40 60 28 80 76 24 20 9 Bra•elton, Method a (Neurologically abno...al at 1!175 high • not 7 yean of age) eta ted Ja-• PHR-IJP!! 44 56 54 60 76 24 40 38 et al., Monitoring 1976 Bcore1 •ul- tiveriate ecoring a high • 51+ low • o-5o Pedrick, No titlea 0,6 39 60 1.8 9 60 0.4 30 0.9 1976 high • 5+ (Pr i•iperae 1 (Pri•iperae, eerly geetetion ...s • 1-4 only) low • 0

OCR for page 149
160 ~ ! . t ..!.- - . . ........"' ~ -=~ ... ,..._ -... ... + ... (; ·- • c- ~~ .. ... ... ...- i; • .. I :!~ .... ........ ...... "' '"'O~.c.a ........ o ... ,..._ I ... - 1 ,..._- 8 c ... ... + : !,; ... 0 ..• ·- • ~ 1 ·- •• ... c- t s ... - I!; • • J ... ... ~ ...- ..,_ I i J ... ... ..,_ + ... - ... "' ... ... j .... .... • c- ~~ ... "' i• ... - .... I ~ 1!" .:~ •• .= ! • .. ... ... .. . .Ill ... i . ... : • . ....- '2': • :;; 1 ....

OCR for page 149
161 the screening parameters of the risk assessment methods reviewed, 1953-1980, and outlined in Table 2. The scoring for risk is described in detail because it often varies--even when one researcher is using the method of another. Por exa.ple, Wilson and Sill (1973) used a high score of 0-40, whereas the originators, Nesbitt and Aubry (1969), used a high score of 0-70. Like- wise, Bebb et al. (1980) used a high score of 4+ when applying the aye- tea designed by Goodwin et al. (1969), who used a high score of 7+. The cbosen cut-off points, i.e., 0-40 or 0-70, 4+ or 7+, affect the propor- tion of the population declared to be at high risk. Thus, the variation between studies in which the same methods were allegedly used diminishes comparability but enhances the possibility of ascertaining which cut-off points are .ast effective (assuming other aspects of study methodology are coaparable). The presentation of high-, medium-, and low-risk inforaation in Table 2 facilitates such comparisons. Scores are provided for middle-level risk, but details of sensi- tivity and false positive rates are given for the highest risk group only. Specificity, false negative rates, and percentage of low-risk waaen experiencing the outcome refer to the lowest risk group examined by the author. The issue of nonconventional birth settings would usually apply to the lowest risk women. Por the sake of simplicity, the middle-risk group is excluded because the values for this group lie between the two extremes. Screening parameters cannot be properly evaluated without informa- tion on the proportion of the group at each level of risk and on the outcome factor (disease). Frequencies in each category affect the accuracy of prediction. The incidence rate is provided for the outcome measures displayed in Table 2, i.e., neonatal complications, low birth weight, and perinatal death. A measure of incidence rather than preva- lence was used because these conditions do not endure as long as condi- tions such as cancer or diabetes. (Incidence rates refer to the number of new cases of a disease in a specific time period for a population at riskJ prevalence rates refer to the number of existing cases of a disease at a specific time for the total population.) The incidence rate is given for the entire population studied irrespective of the level of risk. Predictive value is especially sensitive to the incidence (or prevalence) of the outcome. Because the predictive value is actually the incidence rate of the outcome for a certain risk group, it often airrors the incidence rate in the general population. Where it does not miaic the population rate, there are probably significant differ- ences between that risk group and the expected rate. In Table 2 the predictive value of the low-risk assignment is presented in reverse, that is, the percentage of the low risk group that experiences the undesirable outcome is given because pregnant women assessed as low risk are slated for less exhaustive diagnostic monitoring. Authors varied as to which outcomes they considered important. Specific outcomes and the number of publications in which they were reported are as followss

OCR for page 149
162 Neonatal or perinatal mortality 20 Neonatal morbidity 14 Early gestational age at birth 7 Low Apgar score 7 Low birth weight 6 Intrapartum complications 2 Maternal complications 1 In the studies reviewed, perinatal mortality received the greatest amount of attention. Most of the authors included at least one of the following three outcomesa neonatal complications, low birth weight, and perinatal death (see Table 2). FINDINGS The interplay between incidence of conditions, sensitivity, and predictive value is illustrated in Table 2. Richards and Roberts (1967) point out that if 5 percent of a population is at high risk, the incidence of an outcome must be 16 times higher in the high-risk group than in the low-risk group before the sensitivity of the risk assess- ment can reach 80 percent, the acceptable minimum. The Apgar (1953) score fits this criteria: 6 per'cent of the group was declared high risk, the incidence of perinatal death is 1.2/100, the sensitivity of the score is 92 percent, and 22 percent of the high-risk group died versus 0.1 percent of the low-risk group. The death rate is 22 times higher in the high risk group. Data from Nesbitt and Aubry (1969) indicate that sensitivity is not necessarily influenced by the incidence of the outcome while predictive value clearly is affected. The inci- dence rate of low birth weight is 13 percent and the incidence of perinatal death is 3 percent, but the sensitivity of the high-risk label is the same, 43 percent. Predictive value, on the other hand, varies with incidencer the predictive value of the high-risk assignment with low birth weight as the outcome is 20 percent, and with perinatal death it is 4 percent. The role of sensitivity is to permit correct assignment of women with undesirable outcomes to a high-risk group prior to actual ful- fillment of the outca.e. Thus, the more sensitive a measure, the more often a woman will be correctly identified as high risk. Frequently a trade-off must be made between sensitivity and specificity, an increase in one may yield a decrease in the other. Robel's method had a sensi- tivity of 37 percent with specificity of 51 percent using 126 factors and a dichotomized scoring system. In order to increase sensitivity, a multivariate scoring technique was applied to the data using 39 factorsr the yield was a sensitivity of 94 percent but specificity decreased to 28 percent. The rate of false negatives also decreased from 18 percent to 6 percent, but this represents an improvement for obstetric risk assessaent because fewer women are incorrectly assigned to a low-risk group. Among the methods which used neonatal complications as an outcome,

OCR for page 149
163 the highest sensitivity (94 percent) vas achieved by Robel's method in 1979 with aultivariate scoring, followed by Edwards et al. (75 percent) in 1979 and by Nesbitt and Aubry (47 percent) in 1969. The latter two methods eaplQ¥ed arbitrary weights for the risk factors. There is only a slight indication of correlation between the number of risk factors assessed and the level of sensitivity. Any real association is aasked by the differences in the populations studied. Specificity is increased when fetal monitoring is included. Chik et al. (1979) used Robel's method, plus fetal monitoring, to achieve sensitivity and false negative rates siailar to those Robel's method achieved with aultivariate scoring. However, the specificity is auch higher with the method of Chik et al.: 87 percent versus 28 percent. Indeed, monitoring seems to carry a high specificity with itJ yet, when used alone, it appears to have a high false negative rate. For exa.ple, the method of James et al. (1976) yields a 40 percent false negative rate while Chik et al. (1979), using monitoring and Robel's risk assess- aent scheme, have only a 7 percent false negative rate. Low birth weight vas an outcome variable for very few (six) of the reviewed articles. Only two of them contain enough information to per- ait ca.parisons. Nesbitt and Aubry (1969) attained a sensitivity of 43 percent with their index, which vas applied only at the initial prenatal visit. Later, Aubry and Pennington (1973) added a labor index that in- creased sensitivity (63 percent) and specificity (82 percent) but also increased the rate of false negatives (37 percent). The method of Kaminski et al. (1973), using aultivariate scoring, did not improve on that of Aubry and Pennington (1973). The proportion of women labeled low risk who deliver a low birth weight infant is small when information about labor is included. Risk assessment activities have tended to concentrate on prediction of perinatal or neonatal mortality. Apgar's (1953, 1966) scoring system has the highest level of sensitivity (92 percent) and one of the highest levels of specificity (77 percent) of all the systems reviewed for pre- dicting neonatal mortality. This accuracy is due to the direct observa- tion of the neonate at birth, which is necessary to obtain the Apgar score. The other systems use inforaation available during the prenatal period and attempt to predict perinatal outcome long before the infant appears in the delivery rooa. They are subject to less accuracy than the Apgar score. The methods of Nesbitt and Aubry (1969) and Goodwin et al. (1969) achieved similar sensitivity (43 percent and 46 percent) using prenatal information only. Robel's (1976) systea, which includes the prenatal- through-delivery period, improved the sensitivity to 54 percentJ it also improved the specificity (47 percent). Three studies (Goodwin et al., l969J Bebb et al., 1980J Morrison and Olsen, 1979), using somewhat ca.parable methodology, eaployed the Goodwin et al. instrument but used varying cut-off points to assign high risk. The originators of the scheme achieved the lowest levels of accuracy (sensitivity, 46 percentJ specificity, 28 percent). Their cut-off point of 7+ eliainates false negatives because it vas assigned with knowledge of the level of risk at which no .ore deaths occurred. Morrison and Olsen's (1979) method has a very high false negative rate (30 percent) as well as a high

OCR for page 149
164 level of specificity (70 percent). Examination of the study methods of each investigation suggests the strength of the approach of Hebb et al. (1980) and indicates that its screening parameters are probably the best representation of the method's power. The multivariable instrument of Robel et al. (1973) has been used by several investigators other than the orginators Chik et al. (1979), Sokol et al. (1979, 1980), and Winterset al. (1979). All of these groups have similar study designs. Winters et al. (1979) did not ascertain mortality. Comparing the findings of Sokol et al. (1979) with those of Bobel et al. (1976), it is interesting to note that the screening parameters produced by Sokol and associates are better than Robel's: the sensitivity is 82 percent versus 54 percentJ the false negative rate is 0 percent versus 2 percent. However, as sensitivity increased to 82 percent, the specificity decreased to 31 percent for Sokol's study, compared to 47 percent in Robel's study. The incidence rates for mortality were equal. Sokol et al. (1979) had 26 percent of their group at high/high risk while the similar figure in the Robel et al. (1976) study was 16 percent. Both investigative teams used the same cut-off points for high risk. Probably the two populations studied varied in other ways to account for some of the differences in the parameters. Several authors (Bobel, 1978J Stembera et al., 1975J Winterset al., 1979) have suggested that cut points will have to be determined for each population to which a risk assessment is applied. None of the methods reviewed place many women with perinatal deaths at low risk incorrectly. The predictive value of a low-risk label for subsequent perinatal death is high, i.e., 98 percent of women in the low-risk group have live infants at the end of the perinatal period. Only the method of Donahue and Wan (1973) carries an unacceptable level of risk. Possibly, this is due to the assignment of weights based on prematurity, not mortality, as the outcome variable. Sokol et al. (1977) have noted that inclusion of the intrapartum risk assessment greatly enhances the prediction of perinatal death. Stembera et al. (1975) show similar improvement, rather dramatically. Using only historical and prenatal factors the predictability is 40 percentJ it rises to 73 percent when factors for the entire pregnancy, labor, and neonatal period are included. SCREENING CRITERIA IN UNCONVENTIONAL SETTINGS Risk assessment is used as a screening tool in developing countries to decide where limited resources will be allocated (World Health Organ- ization, 1978). In Great Britain, women were booked for delivery at home, in general practioners' clinics, or in hospitals. The bookings were based on characteristics predisposing to probleas in the d~livery (Butler and Alberman, 1969). Risk factors are assessed in the United States when families desire delivery in unconventional birth settings that do not have major life-supporting equipment immediately available. Use of risk criteria by lay midwives is not well documented in the literature, although personal communication with several midwives suggests that, when criteria are used, they are applied vigorously.

OCR for page 149
165 Bennetts (1982) studied 11 nonhospital childbearing centers admin- istered by certified nurse midwives throughout the United States in 1981 (see Appendix C). Every center has admission criteria baaed partly on risk assessment. Only one center had no written criteria at the tl.e of the study, a risk assessment vas carried out by the cer- tified midwife for admission to the center. The factors assessed by the nonhospital childbearing centers are analogous to those in Table 1, but the centers tended to vary in the degree to which the factors were applied. Five of the centers used absolute criteria for admission, that is, the mere presence of certain factors obviates delivery in the center. The remaining centers em- ployed a risk scoring system similar to those in Tables 1 and 2. A woman labelled low risk and acceptable for delivery at a center may develop ca.plicationa during pregnancy, labor, or delivery, or the neonate may have a problem. Therefore, intrapartum and postpartum transfer criteria exist. These criteria are baaed partly on the resources a center has for handling an emergency and partly on the risk a ca.plication implies for further, more serious outcomes. Bennetts (1982) sampled recorda from each center and obtained information about the proportion of women who began labor in a child- bearing center and were transferred. This transfer rate can serve as an indicator of the predictive value rate of low-risk women to beCome high risk. Thus, it is similar to the rate in Table 2: the percentage of low-risk women who experienced an unfavorable outcome. According to Bennetts' data, 10 percent of women had complications before labor and withdrew from the centers. Another 15 percent of those who began labor in the centers developed complications during or after labor and were transferred. Only one percent of infanta required transfer. The predictive value of low-risk labelling by the time labor begins is 85 percent because 15 percent of women required transfer. Rote that COdMOnly used risk assessment methods are aimed at predicting perinatal mortality and not maternal or intrapartum ca.plicationa. In Table 3 the rates in the childbearing centers compare favorably with tboae of Nesbitt and Aubry (1969), especially for neonatal problema. Of interest is the fact that 14 percent of the transfers from the centers to hospitals experienced no actual complication during labor or delivery (Bennetts, 1982). The datum is testimony to the false positive rate of aa.e of the risk assessment methods used. DISCUSSION AND SUMMARY weakness in Current Methods 1. Only one of the 33 papers presented in Table 1 included informa- tion on the ethnic group of study participants. The major demographic factors of age and ethnic group have been virtually ignored in develop- ing weights fo~ risk factors. Inattention to epidemiological factors probably contributes to the inability to apply the same methods in dif- ferent populations and obtain similar predictability. Separate weight-

OCR for page 149
166 Table 3 Risk of Complications in Pregnancy Percent of Low Birth Weight with Maternal Intrapartua Neonatal Study Complications Complications Complications Bennetts, 1982 10 15 l Nesbitt and Aubry, 1969 24 15 4 Wilson and Sill, 1973 7 (cesarean section only) ing and scoring systeas should be developed for each age group, ethnic group, and social level at least. 2. Past pregnancy history is a large co.ponent of .oat risk assessment •ethods reviewed. These methods are better at screening aultigravada women for risk than primigravida women• e.g., Pedrick's (1976) method for predicting early gestational age bad a sensitivity rate 64 percent higher when applied to •ultiparae women than when it was applied to pri•iparae women. Separate risk assessments need to be developed for primigravida women. 3. Reliability or repeatability of obstetric risk factor assess- ment is rarely addressed in the literature, yet women can be incorrectly labelled through misuse of the instruaent. Robel et al. (1973) include a handbook of definitions and instructions in the way to use their forms. Other authors (Edwards et al., 1979' Baeri, 1974) prefer aiapler approaches. However, reliability in application of screening criteria is especially important when assessment is aade only once in the prenatal period. 4. Expectations for what a risk assessment can do are often •is- aligned with the technique's real capability. In using the risk assess- ment approach, it is important to recall that the risk factors and weights are derived fro. population or grouped data. Most of the accusations of bar• associated with using the risk approach are due to lack of appreciation for the •ecological fallacy• (Parmelee and Baber, 1973' Richards and Roberts, 1967' Wilson and Schifrin, 1980). The risk inherent in any group may not apply to an individual member of the group. The fallacy is that the probabilistic risk of an outcome is assigned to an individual' i.e., whatever is true for the group is supposed to be true for the individual group member. However, the individual may or aay not suffer the undesirable outco.e. It is fair to say, •This person may be at higher risk,• but is incorrect to say, •This person will experience the outooae because of the presence of these risk factors.• The predicted risk of a neonatal death in

OCR for page 149
167 Bobel's high/high risk group is 11 percent. The neonate's actual risk is either 0 percent or 100 percent. The neonate will either survive or dier the neonate will not partially die or be ll percent dead. Thus, even though a woman belongs to a particular group, she may not suffer any undesirable outcome. The reverse is also possibler a woman possessing none of the known risk factors for neonatal death aay lose her infant. In obstetrics, death is a rare event and therein lies the need to exaaine realistically the probability of poor outcomes. The perfor- aance of screening to assign correctly women who will have no problems during pregnancy and child birth to a low-risk group should be seriously considered. From the point of view of alternative birth centers, high false negative rates vis-a-vis perinatal death are anatheaa to the con- cept of women delivering out of hospital. The main risk is two percent or less, and among all women labelled low risk, fewer than 3 per 1,000 will experience a perinatal death. This is much lower than the national perinatal mortality rate. Very little can be gleaned from existing literature on risk assess- .ent methods regarding successful prediction of maternal/intrapartum complications or neonatal morbidity. In general, false negatives are high--usually over 20 percent of women or their infants experiencing undesirable outcomes are incorrectly assigned to a low-risk group. More work must be done in this area. From Bennetts' (1981) study it is also evident that 14 percent of women transferred for predicted co.plications do not experience any ca.plications. At this time, aorbidity in pregnancy and neonates cannot be predicted as accurately as death. Summary Thirty-three risk assessment articles were reviewed in detail, com- prising 19 methods of assigning levels of risk for undesirable outcomes in pregnancy. The predictive power of the methods as screening tools was exaained. Most of the methods are based on the prediction of perinatal death, which hampers their utility for predicting less severe outcomes. Three aajor systems emerge from the literature: (l) the Nesbitt and Aubry (1969) Maternal-child Health Care Index plus the Aubry and Pennington (1973) Labor Index, (2) the Goodwin et al. (1969) Antepartum Fetal Risk Scoring system, and (3) Hobel's (1973) Problem-oriented Risk Assessment system. Numerous other authors proffer their methods, but they are all similar to one of the three major ones. Factors used by each system are similarr it is the importance, or weight, given to each factor and how it is derived that varies. Generally, systems that include prenatal and intrapartum information appear most successful, i.e., strike a good balance between sensitivity and specificity. The cut points for declaring risk level are important and probably should be derived for each population, as should the weights. However, weighting of factors is problematic because we have no •pure• measure of risk--i.e., associating a characteristic with a negative outcome

OCR for page 149
168 prompts intervention, which interrupts the causal chain and distorts the relation of the factor to the outcome. This is as it should be, but predictive power of risk assessment methods will be more inaccurate because of it. REP'EREHCES Alberman, E. D., and H. Goldstein. 1970. The •at-risk• register: A statistical evaluation. British Journal of Preventive and Social Medicine 24:129-135. Ala, H., and T. B. Brazelton. 1975. Comprehensive neonatal assessment (Brazelton Neonatal Behavior Assessment). Birth and the Faaily Journal 2:3-9. Apgar, V. A. 1953. A proposal for a new method of evaluation of the newborn infant. Current Researches in Anesthesia and Analgesia 32:260-267. Apgar, v. 1966. The newborn (Apgar) scoring system. Pediatric Clinics of North America 13:645-650. Aubry, R. H., and J. c. Pennington. 1973. Identification and evaluation of high-risk pregnancy: The perinatal concept. Clinical Obstetrics and Gynecology 16:3-27. Bennetts, A. 1982. Freestanding birth centers. ~ Research Issues in the Assessment of Birth Settings. Washington, D.C.: National Academy Press. Brazelton, T. B. 1973. Neonatal Behavioral Assessment Scale. Philadelphia: Lippincott. Butler, H. R., and E. D. Alberman, eds. Perinatal Proble•: The Second Report of the 1958 Perinatal Mortality Survey. London: Livingstone, Ltd. Chik, L., R. J. Sokol, M. G. Rosen, s. K. Pillay, and s. E. Jarrell. 1979. Trend analysis of intrapartum monitoring data: A basis for a computerized fetal monitor. Clinical Obstetrics and Gynecology 22:665-679. Chng, P. K., M. B. Hall, and I. MacGillivray. 1980. An audit of antenatal care: The value of the first antenatal visit. British Medical Journal 281:1184-1186. Coopland, A. T., L. J. Peddle, T. F. Baskett, R. Rollwagen, A. Si~on, and E. Parker. 1977. A si~lified antepartum high-risk pregnancy scoring form: Statistical analysis of 5459 cases. Canadian Medical Association Journal 116:999-1001. Coradello, H., A. Pollak, s. Scheibenreiter, G. Siabruner, and o. Thalhammer. 1975. A score to pre-calculate the risk of prematurity. In Perinatal Medicine, 4th European Congress of Perinatal Medicine, z. K. Stembera, K. Polacek, and v. Sabata, eds. Acton, Mass.: Publishing Sciences Group. Donahue, c. L., Jr., and T. T. H. Wan. 1973. Measuring obstetric risks of prematurity: A preliminary analysis of neonatal death. American Journal of Obstetrics and Gynecology 116:911-915. Edwards, L. E., M. I. Barrada, R. w. Tatreau, and E. Y. Hakanson. 1979.

OCR for page 149
169 A ai.plified antepartum risk-scoring system. Obstetrics and Gynecology 54:237-240. Bffer, s. B. 1969. Manageaent of high risk pregnancy: Report of a ooabined obstetrical and neonatal intensive care unit. canadian Medical Association Journal 101(7):55-63. Pedrick, J. 1976. Antenatal identification of women at high risk of spontaneous pre-term birth. British Journal of Obstetrics and Gynaecology 83:351-354. Goodwin, J. w., J. T. Dunne, and B. w. Thomas. 1969. Antepartum identification of the fetus at risk. canadian Medical Association Journal 101(8)a57ff. Baeri, A. D., J. South, and J. Naldrett. 1974. A scoring system for identifying pregnant patients with a high risk of perinatal mortality. Journal of Obstetrics and Gynaecology of the British Commonwealth 81:535-538. Halliday, B. L., P. K. Jones, and s. L. Jones. 1980. Method of screening obstetric patients to prevent reproductive wastage. Obstetrics and Gynecology 55:656-661. Bebb, M., I. MacPherson, D. Cudmore, K. Scott, L. Weldon, M. Saart, and B. Ley. 1980. Nova Scotia fetal risk project. canadian Pamily Physician 26al664ff. Robel, c. J. 1976. Recognition of the high-risk pregnant woman. _!n Management of the High-Risk Pregnancy, Williaa N. Spellacy, ed. Balti.ore: University Park Preas. Robel, c. J. 1977. Identification of the patient at risk. !n Perinatal Medicine: Management of the High Risk Petua and Neonate, R. J. Bolognese and R. H. Schwarz, eda. Balti.ore: Williams and Wilkins. Robel, c. J. 1978. Risk assessment in perinatal medicine. Clinical Obstetrics and Gynecology 21:287-295. Robel, c. J. 1979. Aaaeaa. .nt of the high risk fetus. Clinics in Obstetrics and Gynecology 6:367-377. Robel, c. J., M. A. Hyvarinen, D. M. Okada, and w. Oh. 1973. Prenatal and intrapartua high-risk screening: I. Prediction of the high-risk neonate. American Journal of Obstetrics and Gynecology ll7al-9. Robel, c. J., L. Youkelea, and A. Forsythe. 1979. Prenatal and intra- partua high-risk screening: II. Risk factors reassessed. American Journal of Obstetrics and Gynecology 135:1051-1056. James, L., H. R. Rey, R. J. Stark, s. N. caritia, and R. Dwarka. 1976. A new approach to predicting the high-risk neonate: A scoring ayatea for intrapartum .onitoring. In Perinatal Medicine, 5th European Congress of Perinatal Medicine, G. Rooth and L.-B. Bratteby, eda. Stockholma Almqviat and Wikaell International. Kaminski, M., J. Goujard, and c. Ruaeau-Rouquette. 1973. Prediction of low birthweight and prematurity by a multiple regression analysis with aaternal characteristics known since the beginning of the pregnancy. International Journal of Epidemiology 2:195-204. Kessner, D. M., J. Singer, c. E. Kalk, and B. R. Schlesinger. 1973. Infant death: An analysis by maternal ~isk and health care. !n Contrasts in Health Status, Vol. I. Washington, D.C.a National Academy Preas.

OCR for page 149
170 Larks, s. D., and G. G. Larks. 1968. Prenatal prediction of birth process proble•: Biomathematical approaches. Mathematical Bioscience& 3:135-139. Lesinski, J. 1975. High-risk pregnancy: Unresolved proble• of screening, manage•nt, and prognosis. Obstetrics and Gynecology 46:549-603. Morrison, I., and J. Olsen. 1979. Perinatal .ortality and antepartum risk scoring. Obstetrics and Gynecology 53:362-366. Nesbitt, R. B. L., Jr., and R. B. Aubry. 1969. High-risk obstetrics: II. Value of aemiobjective grading system in identifying the vulnerable group. American Journal of Obstetrics and Gynecology 103:972-985. Parmelee, A. H., and A. Baber. 1973. Who is the •risk infant•? Clinical Obstetrics and Gynecology 16:376-387. Prechtl, B. F. R. 1967. Neurological sequelae of prenatal and perinatal complications. British Medical Journal 4:763-767. Rantakallio, P. 1969. Groups at risk in low birthweight infanta and perinatal .ortality. Acta Paediatrica Scandinavica (Supplement Ho. 193). Richards, I. D. G., and c. J. Roberta. 1967. The •at risk• infant. Lancet 2:711-713. Sokol, R. J., M. G. Rosen, J. Stojkov, and L. Chik. 1977. Clinical application of high-risk scoring on an obstetric service. American Journal of Obstetrics and Gynecology 128:652-661. Sokol, R. J., J. Stojkov, and L. Chik. 1979. Maternal-fetal risk aaaeaa. .nt: A clinical guide to monitoring. Clinical Obstetrics and Gynecology 22:547-560. Sokol, R. J., R. B. Woolf, M. G. Rosen, and K. Weingarden. 1980. Risk, antepartum care, and outcome: Iapact of a maternity and infant care project. Obstetrics and Gynecology 56:150-156. Stembera, z. K., J. Zezulakova, J. Dittrichova, and K. Znamenacek. 1975. Identification and quantification of high-risk factors affecting - . fetus and newborn. In Perinatal Medicine, 4th European Congress of Perinatal Medicine, z. K. Stembera, K. Polacek, and v. Sabata, eda. Acton, Mass.: Publishing Sciences Group. Tulchinaky, D. 1980. Use of bioche•ical indices in the aanage•nt of high risk obstetric patients. Clinics on Perinatology 7:413-421. Wilson, R. w., and B. s. Schifrin. 1980. Is any pregnancy low risk? Obstetrics and Gynecology 55:653-656. Wilson, B. w., and B. K. Sill. 1973. Identification of the high risk pregnancy by a scoring system. Hew Zealand Medical Journal 78:437-440. Winters, s., s. Itzkowitz, and K. Johnson. 1979. Prenatal risk assessment: An evaluation of the Robel record in a Mount Sinai clinic population. Mount Sinai Journal of Medicine (Hew York) 46:424-427. World Health Organization. 1978. Risk Approach for Maternal and Child Health Care. WHO Offset Publication Mo. 39, Geneva: WOrld Health Organization. Yeh, s., A. Forsythe, R. I. Lowenaohn, and B. B. Bon. 1977. A study of the relationship between Goodwin's high-risk score and fetal outcome. American Journal of Obstetrics and Gynecology 127:50-55.