This article is commented on by Anderson on pages 202203 of this issue.
Prediction of cognitive abilities at the age of 5 years using developmental follow-up assessments at the age of 2 and 3 years in very preterm children
Article first published online: 21 DEC 2011
© The Authors. Developmental Medicine & Child Neurology © 2011 Mac Keith Press
Developmental Medicine & Child Neurology
Volume 54, Issue 3, pages 240–246, March 2012
How to Cite
POTHARST, E. S., HOUTZAGER, B. A., van SONDEREN, L., TAMMINGA, P., KOK, J. H., LAST, B. F. and van WASSENAER, A. G. (2012), Prediction of cognitive abilities at the age of 5 years using developmental follow-up assessments at the age of 2 and 3 years in very preterm children. Developmental Medicine & Child Neurology, 54: 240–246. doi: 10.1111/j.1469-8749.2011.04181.x
- Issue published online: 10 FEB 2012
- Article first published online: 21 DEC 2011
- Accepted for publication 1st October 2011. Published online 21st December 2011.
Aim This study investigated prediction of separate cognitive abilities at the age of 5 years by cognitive development at the ages of both 2 and 3 years, and the agreement between these measurements, in very preterm children.
Methods Preterm children (n=102; 44 males; 58 females) with a gestational age less than 30 weeks and/or birthweight less than 1000g were assessed at the ages of 2 and 3 years using the second edition of the Bayley Scales of Infant Development, the Child Behaviour Checklist, and a neurological examination, and at the age of 5 years using the third edition of the Wechsler Preschool and Primary Scale of Intelligence.
Results Cognitive development at ages 2 and 3 years explained 44% and 57% respectively of full-scale intelligence at the age of 5 years. Adding psychomotor, neurological, and behavioural outcomes to the regression model could not or only marginally improve the prediction; adding perinatal and sociodemographic characteristics to the regression model increased the explained variance to 57% and 64% respectively. These percentages were comparable for verbal intelligence. Processing speed quotient and especially performance intelligence were predicted less accurately.
Interpretation Not all aspects of intelligence are predicted sufficiently by the Mental Development Index at ages 2 and 3 years. Follow-up of very preterm children until at least the age of 5 years is needed to distinguish between different aspects of cognitive development.
Bayley Scales of Infant Development
Mental Developmental Index
Neonatal intensive care unit
Psychomotor Developmental Index
Processing speed quotient
What this paper adds
- • In very preterm children, early cognitive development is strongly associated with verbal intelligence at the age of 5 years.
- • The prediction of performance intelligence and processing speed by early cognitive development is less accurate.
- • The assessment of early behavioural and neuromotor development add little to the prediction of intelligence.
Follow-up assessments of very preterm children are necessary in order to guarantee continued care and evaluation of treatments during the perinatal period.1 Because follow-up rates are highest in the first years after the neonatal intensive care unit (NICU) period and because follow-up is expensive and time-consuming, outcomes at 18 to 24 months of age are frequently reported as an end point of neurodevelopmental follow-up.2 The content and timing of follow-up have to be attuned as much as possible to the problems and disabilities often seen in very preterm children.3
Outcomes of several follow-up studies suggest that evaluation at 18 to 24 months of age should rather be seen as an intermediate point.4,5 Developmental disabilities become fully apparent only when the child starts to develop complex functions that can be assessed.6 Roberts et al.4 reported poor agreement between assessments at 24 months and 8 years, especially in the cognitive domain (κ=0.11), in their cohort of extremely preterm children. Hack et al.5 also found a poor predictive validity of cognitive development at the age of 20 months in a sample born preterm. Agreement appears to be better in predicting the outcome of children with severe disabilities.7
From preschool age onwards, different cognitive abilities can be distinguished.8 Moreover, studies reveal that, in addition to global cognitive disabilities, specific learning deficits are present in preterm infants.6 To our knowledge, the only study that reports on the prediction of specific aspects by early cognitive development is one by Crowe et al.,9 in which a more heterogeneous group was studied and older versions of developmental tests were used. Crowe et al. concluded that cognitive abilities can be predicted only to a small extent.
This study focuses on the prediction of distinct cognitive abilities at the age of 5 years from the viewpoint of early follow-up in very preterm children with a gestational age of less than 30 weeks and/or a birthweight of less than 1000g who were able to complete an intelligence test. Other studies have already shown that cognitive assessment at the age of 2 and 3 years does not predict later cognitive functioning well enough. In the current study, we investigate to what extent cognitive functioning at the age of both 2 and 3 years predicts not only full-scale but also verbal and performance intelligence and processing speed at the age of 5 years and what the agreement between these measurements is. Because it is harder to distinguish between different developmental domains the younger a child is, we hypothesize that domains of development other than cognition at the age of 2 and 3 years (psychomotor, neurological, and behaviour) will improve the prediction of cognitive development at the age of 5 years. It is not clear whether perinatal morbidities already have their full impact on development at the age of 2 and 3 years, or, alternatively, whether they still have an additional impact on outcomes at the age of 5 years.10 We therefore aim to explore whether perinatal morbidities improve the prediction of cognitive outcomes at the age of 5 years when development at the ages of 2 or 3 years is already accounted for, and which consequences of perinatal events manifest themselves after the age of 2 and 3 years. Finally, sociodemographic factors tend to become more important as children grow up;6 additional effects of these factors will therefore be investigated.
This is a single-centre longitudinal cohort study. In the Netherlands, NICU care is centralized in only 10 specialized hospitals.
The study sample consisted of a cohort of 5-year-old very preterm children (n=102; 44 males; 58 females) for whom earlier follow-up data were available. Children born before 30 weeks’ gestation and/or with birthweights below 1000g, and who reached the corrected age of 5 years between December 2007 and June 2009 were included. Other inclusion criteria were (1) hospitalization in the NICU of our hospital, (2) the availability of at least one developmental follow-up score at the age of 2 and/or 3 years, and (3) residency in the Netherlands. Exclusion criteria were (1) participation in one of two other studies11,12 because of the use of different instruments and different timing of follow-up, (2) having a genetic syndrome, or (3) being too disabled to be assessed. Disability was defined by a considerable cognitive delay (Mental Developmental Index [MDI] score of below 55 at the age of 2 and/or 3y when age was corrected for prematurity), cerebral palsy with the inability of the child to point to pictures, or severe hearing problems making regular intelligence assessment impossible.
Ethical approval of this study was given by the Medical Ethical Committee of the Academic Medical Centre (Amsterdam, the Netherlands). Informed consent to the research and publication of results was given by parents of all children. At the age of 5 years, the child’s intelligence was assessed by a trained child psychologist. Data concerning the child’s functioning at 2 and 3 years of age (neurological, cognitive, and psychomotor abilities, as assessed by a trained paediatrician and child psychologist, and behaviour measured with questionnaires completed by parents) were available. For reasons of comparability over time,4,13 all scores were calculated on the basis of the child’s corrected age. Cognitive scores were considered mildly or severely abnormal if they were more than 1 or 2SD below the mean respectively.
Intelligence at the age of 5 years was assessed using the Dutch standardization of the third edition of the Wechsler Preschool and Primary Scale of Intelligence.14 Four composite scores were calculated: Full-scale IQ, Verbal IQ, Performance IQ, and processing speed quotient (from now on referred to as FSIQ-5, VIQ-5, PIQ-5, and PSQ-5 respectively).
The Dutch standardization of the Bayley Scales of Infant Development Second Edition (BSID-II-NL)15 was used to assess MDI and the Psychomotor Development Index (PDI) at the ages of 2 and 3 years. MDI and PDI at the ages of 2 and 3 years will be referred to as MDI-2, PDI-2, MDI-3, and PDI-3 respectively.
Emotional and behavioural problems were assessed at the ages of 2 and 3 years using the Dutch version of the Child Behaviour Checklist (CBCL 1.5–5).16 The total problem T score was used.
Neurological development was assessed at ages 2 and 3 years using the Hempel neurological examination.17 Neurological development was classified as normal, suspect, or abnormal.
Perinatal data were abstracted from an ongoing prospective database used for all infants admitted to our NICU. The level of parental education was used as a measure of socioeconomic status. Parents who attended the lowest type of college or less were rated ‘low level of education’ (total years post-elementary schooling <6). Parents who graduated from a middle-level college were rated with ‘middle level of education’ (total years post-elementary schooling 6–8). Parents who attended the highest-level college or university were rated with ‘high level of education’ (total years post-elementary schooling >8). The combined parental education score was low if one or both parents had a low level of education, middle if both parents had a middle-level education, and high if one or both parents had a high-level education.
Univariate analyses were carried out to study differences in baseline characteristics between participants, excluded children, and non-participants. Mean differences between measurement occasions were tested using repeated measures analysis of variance and post-hoc paired samples t-tests. The agreement between MDI at the age of 2 and 3 years and cognitive measures at the age of 5 years was determined with a linearly weighted kappa coefficient. Categories were interpreted as proposed by Landis and Koch18 (0<κ<0.2, slight; 0.2<κ<0.4, fair; 0.4<κ<0.6, moderate; 0.6<κ<0.8, substantial; 0.8<κ<1, almost perfect). Hierarchical regression analyses were performed to predict the continuous scores of FSIQ-5, VIQ-5, PIQ-5, and PSQ-5. In each step, an independent variable or group of variables was added. If the new model added significantly (p<0.05) to the explained variance of the dependent variable, the new model was accepted and the next step was analysed. MDI was entered in the first step. In the following five steps, PDI, neurological examination, behaviour, and perinatal and sociodemographic characteristics were added consecutively. Only perinatal characteristics that were significantly associated with the dependent variable in univariate analyses (p<0.10) were added. Via a backwards procedure, perinatal characteristics were removed from the model until it was significant with a p-value of less than 0.05 and the individual variables had p-values of less than 0.10. For sociodemographic characteristics, the same procedure was followed. Regression analyses were performed separately for the assessments performed at 2 and 3 years of age.
The participants were children born before 30 weeks’ gestation and/or with a birthweight below 1000g who turned 5 years old in the period of recruitment. Thirty-seven children were excluded: 16 had participated in another study, nine children had no developmental score available from an earlier follow-up, four children were too disabled, three had a genetic syndrome, and five families had moved abroad. However, 113 children remained eligible and were invited to participate. Eleven children were lost to follow-up or because their parents refused participation. With respect to baseline characteristics, participants (n=102) received indometacin more often than non-participants and excluded children (n=48; p=0.02; Table I).
|Participants (n=102)||Excluded children and non-participants (n=48)||p-value|
|Perinatal and child characteristics|
|Gestational age (wks + d), mean (SD)||28+5/7 (1 + 4/7)||28+6/7 (1+5/7)||0.58|
|Birthweight (g), mean (SD)||1040 (253)||1093 (260)||0.23|
|Sex: male (%)||44 (43.1)||27 (56.3)||0.13|
|Gestational age ≥30wks and birthweight <1000g (%)||14 (13.7)||5 (10.4)||0.57|
|Very small for gestational age (birthweight <p 2.3 for gestational age) (%)||15 (14.7)||8 (16.7)||0.76|
|Postnatal dexamethasone (%)||4 (3.9)||4 (8.3)||0.44|
|Indometacin for patent ductus (%)||28 (27.5)||5 (10.4)||0.02a|
|Oxygen support at 36wks pma (%)||15 (14.7)||5 (10.4)||0.47|
|Sepsis and meningitis (%)||28 (27.5)||16 (36.4)||0.46|
|Necrotizing enterocolitis, stage 2 and 3 (%)||2 (2.0)||1 (2.1)||1.00|
|PVL 1b (%)||5 (4.9)||4 (8.3)||0.47|
|Subependymal haemorrhage (%)||25 (24.5)||13 (27.1)||0.74|
|IVH 2 (%)||3 (2.9)||2 (4.2)||1.00|
|IVH 3b (IVH 3) (%)||3 (2.9)||0 (−)||0.55|
|Posthaemorrhagic hydrocephalus (4mm >p 97 of Levene curves) (%)||4 (3.9)||0 (−)||0.31|
|Severely abnormal ultrasound (PVL 2–4, IVH 3–4, and/or PHH) (%)||4 (3.9)||0 (−)||0.31|
|Cerebral palsy at the age of 5y||5 (4.9)||4 (10.5)c||0.23|
|Parental foreign country of birthd (%)||31 (30.4%)||–||–|
|Parental education at child’s age of 5y|
|High (%)||44 (43.1%)||–||–|
|Low–middle (%)||58 (56.9%)||–||–|
Of 102 participating children, 101 children were assessed at age 2 years and 96 at age 3 years. MDI, PDI, neurological examination, and CBCL scores at the age of 2 years were obtained for 100, 100, 102, and 95 children respectively, and at the age of 3 years for 96, 94, 94, and 90 children respectively. The Dutch standardization of the third edition of the Wechsler Preschool and Primary Scale of Intelligence scores were obtained for all children.
Course of development
Table II displays mean scores on developmental tests and the proportions of children with abnormal scores. Significant differences were found between MDI-2, MDI-3, and FSIQ-5 (p=0.000). MDI-3 scores were on average 11 and 9 points higher than MDI-2 and FSIQ-5 respectively (p=0.000).
|Corrected age of 2y, n=101|
|MDI (BSID-II), n=100|
|Mean (SD)||91 (18)|
|Mildly abnormal (%)||20 (20.0)|
|Severely abnormal (%)||13 (13.0)|
|Corrected age of 3y, n=96|
|MDI (BSID-II), n=96|
|Mean (SD)||102 (14)|
|Mildly abnormal (%)||11 (11.5)|
|Severely abnormal (%)||3 (3.1)|
|Corrected age of 5y, n=102|
|FSIQ (WPPSI-III), n=102|
|Mean (SD)||93 (17)|
|Mildly abnormal (%)||16 (15.7)|
|Severely abnormal (%)||10 (9.8)|
|VIQ (WPPSI-III), n=102|
|Mean (SD)||95 (17)|
|Mildly abnormal (%)||21 (20.6)|
|Severely abnormal (%)||9 (8.8)|
|PIQ (WPPSI-III), n=102|
|Mean (SD)||93 (15)|
|Mildly abnormal (%)||19 (18.6)|
|Severely abnormal (%)||8 (7.8)|
|PSQ (WPPSI-III), n=102|
|Mean (SD)||93 (18)|
|Mildly abnormal (%)||19 (18.6)|
|Severely abnormal (%)||12 (11.8)|
|All IQ scores|
|Normal (%)||56 (54.9)|
|At least one abnormal IQ score (mild or severe) (%)||46 (45.1)|
|At least one severely abnormal IQ score (%)||17 (16.7)|
Agreement in cognitive scores
Figure 1 shows the correlation between MDI-2 and FSIQ-5. Children with an MDI-2 of 97 or higher all had FSIQ-5 scores in the normal range. Only in children with an MDI-2 score of 118 or higher (n=5) did 100% have scores in the normal range on all 5-year cognitive outcomes.
The correlation coefficient (r) between MDI-2 and FSIQ-5 (0.715), and MDI-3 and FSIQ-5 (0.737) was not statistically different. Appendix I displays the number of children with abnormal MDI-2/3 and FSIQ-5 scores and the movement of children across categories over time.
Agreement between MDI and FSIQ was moderate (κ=0.48 for MDI-2 and 0.40 for MDI-3). Of all different cognitive abilities measured at age 5 years, MDI-2 had the highest agreement with VIQ-5 (κ=0.63). PIQ-5 and PSQ-5 showed lower agreement with MDI-2 (κ=0.36 and 0.47 respectively). Agreement of MDI-3 with VIQ, PIQ, and PSQ was 0.42, 0.41, and 0.37 respectively.
Prediction of 5-year outcome
Table III displays the prediction models for cognitive outcomes at the age of 5 years. MDI-2 and MDI-3 were positively associated with FSIQ-5 (44% and 56% explained variance respectively). PDI-2, PDI-3, and neurological functioning did not improve the prediction of FSIQ-5, with MDI already in the model. Adding behaviour problems at the age of 2 years somewhat improved the prediction. Neonatal sepsis and/or meningitis in both models and postnatal dexamethasone in the MDI-3 model increased the amount of explained variance and were negatively associated with FSIQ-5. Middle/low parental education and foreign parental country of birth were negatively associated with FSIQ-5 and improved the amount of explained variance in the final step. A total of 57% and 64% of the variance of FSIQ-5 was explained in the MDI-2 and MDI-3 model respectively.
|Developmental assessment at||Developmental assessment at||Developmental assessment at||Developmental assessment at|
|2y (n=94)||3y (n=96)||2y (n=100)||3y (n=96)||2y (n=100)||3y (n=96)||2y (n=100)||3y (n=94)|
|Step 1: cognitive development (MDI)||0.662a||0.747a||0.734a||0.765a||0.490a||0.530a||0.547a||0.623a|
|Step 2: psychomotor development (PDI)||–1||–1||–1||–1||–1||–1||0.254b||0.296a|
|R2 (ΔR2)||0.34 (0.04)||0.44 (0.05)|
|Step 3: neurological functioning||–1||–1||–1||–1||–1||–1||−0.173b||–1|
|R2 (ΔR2)||0.36 (0.03)|
|Step 4: behaviour||−0.190b||–1||–1||–1||–1||–1||–1||–1|
|R2 (ΔR2)||0.47 (0.03)|
|Step 5: perinatal characteristics||–1||–1|
|Very small for gestational age||–2||–2||–2||−0.109||–2||–2|
|Oxygen support at 36wks postmenstrual age||–2||–2||–2||–2||–2||–2|
|Sepsis and meningitis||−0.195*||−0.122||−0.176b||−0.143b||–0.153||–2|
|Severely abnormal ultrasound||–2||–2||–2||–2||–2|
|R2 (ΔR2)||0.51 (0.04)||0.59 (0.03)||0.57 0(0.03)||0.62 (0.03)||–2||0.43 (0.07)||0.52 (0.08)|
|Step 6: sociodemographic characteristics||–1||–1||–1|
|Parental foreign country of birth||−0.137||−0.205a||−0.157b||−0.194a||−0.167|
|R2 (ΔR2)||−0.57 (−0.06)||0.64 (0.05)||0.62 (0.05)||0.65 (0.03)||0.31 (0.07)|
MDI-2 and MDI-3 were positively associated with VIQ-5 (54% and 59% explained variance respectively). Psychomotor development, neurological functioning, and behaviour problems did not improve the prediction of VIQ-5. Neonatal sepsis and/or meningitis in both models and being very small for gestational age in the MDI-3 model added to the explained variance and were negatively associated with VIQ-5. Sociodemographic characteristics that were subsequently negatively associated with VIQ-5 were foreign parental country of birth and, only in the 2-year model, middle/low parental education.
MDI-2 and MDI-3 were positively associated with PIQ-5 (24% and 28% explained variance respectively). Middle/low parental education and foreign country of birth increased the amount of explained variance of PIQ-5 in the 2-year model only.
MDI-2 and MDI-3 were positively associated with PSQ-5 (30% and 39% explained variance respectively). With MDI in the model, PDI-2 and PDI-3 were also positively associated with PSQ-5. An abnormal neurological outcome at the age of 2 years was negatively associated with PSQ-5. Perinatal and child characteristics that improved the amount of explained variance and were negatively associated with PSQ-5 were male sex in both models, sepsis and/or meningitis in the MDI-2 model, and postnatal dexamethasone in the MDI-3 model.
In this study, the association between cognitive development at the age of 2 and 3 years and full-scale, verbal, and performance intelligence and processing speed at the age of 5 years was investigated in very preterm children who were able to complete an intelligence test. Agreement between cognitive development at the ages of 2 and 3 years and cognitive functioning at the age of 5 years was moderate to substantial for full-scale and verbal intelligence and fair to moderate for performance intelligence and processing speed.
We found a relatively high mean cognitive score at the age of 3 years compared with the scores at ages 2 and 5 years (all corrected for gestational age). Apparently, the relatively high cognitive scores we found in an earlier study at age 3 years13 did not persist until the age of 5 years. The question arises whether the differences in mean test scores represent true development or whether the favourable outcome at age 3 years is an artefact of the test. General stability of developmental scores19–21 lends support to the latter in the absence of a theoretical foundation supporting the specifically high developmental scores at age 3 years in preterm children. Regarding the preferability of test age (2 or 3y), the results of our study are not univocal. Agreement and correlation statistics point to different preferred test ages, but the differences are too small to make a well-founded recommendation. In spite of this, the age of 2 years appears to be preferable because (1) the number of false negatives will be much lower, and (2) priority should be given to starting intervention at the earliest possible age in children with developmental problems.
Research has shown that agreement between the BSID-II MDI and later intelligence is poor and the predictive validity of MDI is low.4,5 In the study by Crowe et al.,9 in which children were followed up after NICU treatment at the ages of 2 years and 4 years 6 months, the explained variance of verbal and performance intelligence by MDI (first edition of the BSID) was 21% and 17% respectively. The current study reports higher proportions of explained variance, especially for verbal intelligence. An explanation for this difference may be the use of different versions of the BSID. A substantial part of the cognitive scale of the second edition consists of items that require verbal comprehension or responses.
Performance intelligence and processing speed predominantly require non-verbal responses. It is not surprising, therefore, that only one-quarter of the variance of performance intelligence was explained by MDI. The prediction of performance intelligence was not improved by other domains of infant development. In our study group, abnormal scores on performance intelligence and processing speed were obtained equally frequent as abnormal scores on verbal intelligence. A study by Mulder et al.22 showed that deficits in processing speed explained significant group differences in academic attainment between 9- to 10-year-old very preterm and term children (in middle childhood). The present study results add to this in that early developmental assessments only marginally predict performance and processing speed development. Therefore, follow-up timed at the age of 2 and 3 years is insufficient to reveal the full impact of preterm birth on all affected areas of cognitive development. In line with the above, we also found that only children with an MDI-2 of 118 or higher had normal scores (≥85) on every aspect of intelligence at the age of 5 years. Moreover, almost half of the children in our study group had a mildly abnormal score for one of the areas of intelligence at age 5 years, while at the ages of 2 and 3 years only 33% and 15% respectively, had a mildly abnormal MDI score. Note that children who were too disabled to be assessed are not included in these percentages.
In this study, it was shown that developmental domains other than cognition at the ages of 2 and 3 years could not or could only marginally improve the prediction of cognitive functioning at the age of 5 years. The hypothesis that a broad developmental assessment can be used to substantially contribute to a prediction of cognition at a later age was not confirmed.
A study by the National Institute of Child Health and Human Development Neonatal Research Network23 in infants with an extremely low birthweight showed that neonatal infections are associated with poor neurodevelopmental outcomes at the age of 18 to 22 months. The present study demonstrates that the age of 2 years is too early to show the full impact of infections. As complex cognitive functions develop until young adulthood,24 longer-term follow-up might reveal further impact of perinatal events.
In an earlier paper on outcomes in this cohort,25 we described the interaction of the level of parental education with preterm birth on cognitive development at the age of 5 years. In the present study it was, as expected, shown that sociodemographic factors have additional impact on the 5-year outcome even if MDI at the ages of 2 and/or 3 years is already accounted for.
Strengths of the present study were the focus on different cognitive abilities and the availability of longitudinal developmental assessment at three time points. One limitation of the study was the lack of a term-born comparison group. Another limitation was the exclusion of children who were too disabled to be assessed. Results cannot be generalized to all low-functioning children. As the third version of the BSID is increasingly being used, the predictive value of that version also requires further examination. Nevertheless, many papers are still being published using the BSID-II as main outcomes.2 This study can be of help in interpreting these studies with respect to later outcomes.
In a cohort of very preterm children without severe disabilities, early cognitive development predicted verbal intelligence more accurately than performance intelligence and processing speed. Follow-up of very preterm children until at least the age of 5 years is needed to distinguish between different aspects of cognitive development and to find the effects of interventions aiming to improve cognitive outcomes.
- 3American Academy of Pediatrics. Follow-up care of high-risk infants. Pediatrics 2004; 114: 1377–97.
- 8Wechsler Preschool and Primary Scale of Intelligence (WPPSI-III), 3rd edn. San Antonio, TX: Psychological Corporation, 2002..
- 14WPPSI-III-NL Wechsler Preschool and Primary Scale of Intelligence, 3rd edn. Nederlandse bewerking. Amsterdam: Pearson Assessment and Information B.V., 2009., .
- 15Bayley Scales of Infant Development, 2nd edn. Dutch version. Lisse (NL): Swets, 2002., , , .
- 16Praktische Handleiding voor de CBCL. Assen: Van Gorcum, 1990., , , .
|Normal FSIQ||Mildly abnormal FSIQ||Severely abnormal FSIQ||Total|
|Normal MDI||61 (61.0)||6 (6.0)||0 (0)||67 (67)|
|Mildly abnormal MDI||13 (13.0)||3 (3.0)||4 (4.0)||20 (20.0)|
|Severely abnormal MDI||2 (2.0)||5 (5.0)||6 (6.0)||13 (13.0)|
|Total||76 (76.0)||14 (14.0)||10 (10.0)||100 (100)|
|Normal MDI||69 (71.9)||11 (11.5)||2 (2.1)||82 (85.4)|
|Mildly abnormal MDI||4 (4.2)||2 (2.1)||5 (5.2)||11 (11.5)|
|Severely abnormal MDI||0 (0)||1 (1.0)||2 (2.1)||3 (3.1)|
|Total||73 (76.0)||14 (14.6)||9 (9.4)||96 (100)|