Validation of Neurologic Impairment Diagnosis Codes as Signifying Documented Functional Impairment in Hospitalized Children

Abstract

Objective:

To assess the performance of previously published high-intensity neurologic impairment (NI) diagnosis codes in identification of hospitalized children with clinical NI.

Methods:

Retrospective study of 500 randomly selected discharges in 2019 from a freestanding children’s hospital. All charts were reviewed for 1) NI discharge diagnosis codes and 2) documentation of clinical NI (a neurologic diagnosis and indication of functional impairment like medical technology). Test statistics of clinical NI were calculated for discharges with and without an NI diagnosis code. A sensitivity analysis varied the threshold for “substantial functional impairment.” Secondary analyses evaluated misclassified discharges and a more stringent definition for NI.

Results:

Diagnosis codes identified clinically documented NI with 88.1% (95% CI: 84.7, 91) specificity, and 79.4% (95% CI: 67.3, 88.5) sensitivity; NPV was 96.7% (95% CI: 94.8, 98.0), and PPV was 49% (95% CI: 42, 56.1). Including children with milder functional impairment (lower threshold) resulted in NPV of 95.7% and PPV of 77.5%. Restricting to children with more severe functional impairment (higher threshold) resulted in NPV of 98.2% and PPV of 44.1%. Misclassification was primarily due to inclusion of children without functional impairments. A more stringent NI definition including diagnosis codes for NI and feeding tubes had a specificity of 98.4% (95% CI: 96.7-99.3) and sensitivity of 28.6% (19.4-41.3).

Conclusions:

All scenarios evaluated demonstrated high NPV and low-to-moderate PPV of the diagnostic code list. To maximize clinical utility, NI diagnosis codes should be used with strategies to mitigate the risk of misclassification.

Introduction

Children with severe neurological impairment (NI) are a growing¹ population of children with complex medical needs.² Children with NI have common clinical findings (for example, spasticity and gastroesophageal reflux), but divergent underlying diagnoses (for example, cerebral palsy, adrenoleukodystrophy, and trisomy 18).³ Children with NI have high rates of healthcare hospitalization—0.2% of children² account for 20% of bed-days in pediatric hospitals⁴ They have more frequent⁵ and longer⁶ inpatient stays, increased critical care admissions⁷ and higher use of home care² than other children. This substantial medical support often leads to challenging decisions for their families.⁸ Health administrative data provide an opportunity to help fill the evidence gap for children with NI,⁹ but the accuracy of methods used to identify children with NI in large data sets is unknown.

Importantly, children with NI not only have an underlying neurologic or genetic diagnosis but also functional impairments⁴—like developmental delays,⁴ use of indwelling medical devices (such as feeding or tracheostomy tubes),¹⁰ and need for assistance with activities of daily living¹¹—that contribute to their medical complexity.¹⁰ Functional impairment distinguishes NI from the broader category of neurologic disorders,⁴ but is not explicitly included in most diagnosis code classification schemes.^{4, 12} For some diagnoses, like cerebral palsy or trisomy 18, functional impairment is a consistent feature of the underlying disorder, although the degree of impairment can vary substantially.^{13, 14} Other diagnoses like hypoxic-ischemic encephalopathy or genetic seizure syndromes are associated with varying clinical outcomes: affected children have a neurologic disorder, but only some children manifest the functional impairments consistent with NI.^{15, 16}

The methodology typically used in health services research identifies NI using hospitalization discharge diagnoses. Researchers have developed lists of diagnosis codes consistent with NI.^{4, 12, 17} Although these NI codes have been widely for research and clinical initiatives,^{3, 17–20} they have not undergone testing to determine the likelihood of functional impairment among children with the designated diagnoses, which is critical to design and interpretation of research studies using them. Validation of health administrative data algorithms to minimize the risk of misclassification bias has been identified as a priority by an international consortium of health service researchers.²¹ As the NI code lists are increasingly being used to address clinical questions,^{18, 19} understanding the population they identify is crucial to appropriately applying the evidence at the bedside.

The study sought to assess the performance of NI diagnosis codes¹² in identification of hospitalized children with NI, compared to documentation in the health record of NI with substantial functional impairment. We hypothesized that the NI diagnosis codes would have high sensitivity and lower specificity.

Methods

Study site and population

This retrospective study was conducted at the Hospital for Sick Children, a 350-bed tertiary care pediatric hospital in Toronto, Canada with approximately 16,000 inpatient admissions per year. We evaluated a subset of 500 randomly selected inpatient admissions to evaluate an alternative hypothesis that the Cohen’s kappa in identification of high-intensity NI between expert review and the code list would be greater than 0.7, assuming a prevalence of 13.5%⁴ and baseline kappa of 0.82 (80% power, 0.05 alpha). All inpatient admissions at the Hospital for Sick Children with dates of discharge between January 1, 2019 and December 31, 2019 were included on a list of encoded admissions. We used a random number generator to select 500 admissions from the sampling frame.

Chart abstraction and case ascertainment

Trained research staff (V.C.) abstracted demographic data and reviewed the 500 charts for evidence of an NI-associated neurologic/genetic diagnosis and an indicator of functional impairment⁴ which we operationalized as including ≥1 of: 1. developmental delay, 2. mobility devices, 3. assistance with activities of daily living, or 4. medical technology. She classified charts with an applicable diagnosis and functional impairment as “NI,” charts with neither feature as “no NI,” and all others “Unsure.” Expert reviewers (K.N. and C.D.)—both pediatricians with clinical and research expertise in NI—assessed a training sample (20%) of the first 100 “NI”/“no NI” charts.²² For the “unsure” charts, both expert reviewers assessed a training sample (25 charts), then one expert reviewer (K.N.) assessed the rest with input from C.D. as needed. Finally, “unsure” charts were reviewed to identify children with milder degrees of functional impairment; these charts were labeled “less severe impairment.”

NI code list identification

The American and Canadian versions of ICD-10 have minor differences based on expansion of sub-chapters (for example, the Canadian version only lists the subchapter E72.5, but the American version includes 5 subcodes E72.50-E72.59). American ICD-10 discharge diagnosis codes on the high-intensity NI code list¹² were translated to the Canadian version of ICD-10 by K.N. and J.F. with assistance from J.T., all of whom have clinical and research expertise with NI. (List is available by request). All ICD-10 discharge diagnosis associated with the index admission and with admissions occurring up to 2 years prior to the index admission date (lead-in period) were abstracted in order to capture chronic neurologic impairment that might be missed on a single hospitalization record. Discharge diagnoses that appeared on the high-intensity NI diagnosis code list¹² were identified, and children with ≥1 NI diagnosis codes were classified “NI.” Children without admissions including NI diagnoses were classified as “no NI.” The research team was blinded to the diagnosis code classification during chart review.

Primary Analysis

We described clinical characteristics of the study population. To evaluate chart review inter-rater agreement, we calculated Cohen’s kappa coefficients. Using the chart review as the gold standard, we calculated test statistics (sensitivity, specificity, negative predictive value [NPV] and positive predictive value [PPV]) to evaluate performance of classification by NI diagnosis codes. A Cohen’s kappa coefficient comparing presence of NI on expert chart review and by NI diagnosis codes was also calculated. Because this study evaluated performance in a children’s hospital where prevalence of NI is likely much higher (in one study, NI prevalence was 13.5% in children’s hospitals vs. 2.5% in community hospitals⁴) we also estimated PPV and NPV for lower prevalence scenarios (2.5% and 5%). Recognizing the inherent challenges of binary classification for cases with less severe functional impairment, we performed a sensitivity analysis varying the threshold for NI classification by assigning all “less severe impairment” charts as “NI” (lowest threshold) and as “not NI” (highest threshold). Analyses were completed using R version 3.6.2.

Secondary Analyses

We re-evaluated misclassified charts (false positives and false negatives) to characterize the reasons for misclassification. We assessed misclassified charts to ensure that charts with less severe impairment were flagged for sensitivity analysis. To evaluate potential markers of functional impairment in health administrative data for inclusion in an algorithm to identify NI, we described cohort characteristics available in discharge records (for example, NI ICD-10 diagnosis category, non-NI complex chronic conditions, technology dependence) stratified by clinical NI categorization (NI, less severe impairment, not NI). We defined complex chronic conditions²³ and technology dependence² based on a previously published ICD-10 codes. The study was neither designed nor powered to statistically evaluate these differences. We also described test statistics for a more restrictive method to identify NI, requiring ICD-10 codes for both NI and a feeding tube.

Ethics Statement and Guideline Use

Ethics approval was obtained from the Research Ethics Board at the Hospital for Sick Children. (#100006842). We followed the Standards for Reporting of Diagnostic accuracy (STARD) guidelines in manuscript preparation.²⁴ Protocol available on request.

Results

Study Population

Children in the cohort had a median age of 6.0 years (interquartile range [IQR] 1.5, 12.3) and median length of stay of 2.6 days (IQR: 1.1, 6.0). (Table 1) The most common discharging services were Surgery (n=160, 32.0%) and General Pediatrics (n=105, 21.0%). A functional impairment was documented in 98 charts (19.6%), most commonly a form of medical technology (n=65, 13.0%).

Table 1.

Cohort Characteristics

Number of discharges	500
Age category, n (%)*
≤1 year	148 (29.6%)
2-5 years	88 (17.6%)
6-11 years	124 (24.8%)
≥12 years	140 (28.0%)
Sex, n (%)
Female	227 (45.4%)
Male	273 (54.6%)
Length of stay, n (%)
<1 day	98 (19.6%)
1-2 days	177 (35.4%)
3-6 days	116 (23.2%)
≥7 days	109 (21.8%)
Discharging service, n (%)
Surgery and non-cardiac surgical subspecialties	160 (32.0%)
General pediatrics	105 (21.0%)
Oncology	53 (10.6%)
Cardiology/cardiac surgery	37 (7.4%)
Neurology	34 (6.8%)
Neonatology	32 (6.4%)
Gastroenterology	27 (5.4%)
Psychiatry	17 (3.4%)
Nephrology	17 (3.4%)
Other medical subspecialty	10 (2.0%)
Critical care	8 (1.6%)
Functional impairment type,**n (%)
None	402 (80.4%)
Medical Technology	65 (13.0%)
Developmental Delay	44 (8.8%)
Assistance with ADLs	4 (0.8%)
Mobility Devices	2 (0.4%)

^*Charts were selected at random and categorized into clinically relevant age groups.

^**May have more than 1.

Reviewer Agreement

Of the 500 charts reviewed, 437 (91%) were easy to classify (“NI”/“No NI”). In the training sample of 22 “easy to classify” charts, the reviewers exhibited a 3-rater Fleiss’ Kappa of 0.81 (p<0.001). For the remaining 47 (9%) “unsure” charts, the Cohen’s Kappa for the two pediatricians was 0.57 (95% CI: 0.29-0.86) on the 33 charts assessed by both. In all cases, agreement was reached through discussion. Because our sampling frame included all 2019 hospital discharges, 13 children had 2 hospitalizations reviewed, and the same NI status was assessed for both discharges in 12 cases (92%).

Characteristics of Children with NI

Children with NI accounted for 788 hospital days (22.4%) of the total 3513 hospital days. They had a median age of 5.0 years (IQR 1.4, 11.0), and 29 (46.0%) were female. Their most common discharging services were general pediatrics (n=22, 34.9%), surgery (n=16, 25.4%) and neurology (n=13, 20.6%). A majority of children with NI were assisted by medical technology (n=38, 60.3%) and had developmental delays described in their charts (n=34, 54%). The most common underlying diagnoses associated with NI were genetic (for example, specific point mutations and eponymous syndromes; n=21, 33.3%) and cerebral palsy (n=11, 17.5%). (Table 2)

Table 2.

Primary Cause of Neurologic Impairment (NI) from Chart Review.

Total	63
Rare genetic syndromes*	21 (33.3%)
Cerebral palsy	11 (17.5%)
Trisomy 21	7 (11.1%)
Underlying NI diagnosis not listed†	7 (11.1%)
Epilepsy	4 (6.3%)
Stroke	4 (6.3%)
Metabolic disorder	3 (4.8%)
Neuromuscular disorders	2 (3.2%)
Spina bifida	2 (3.2%)
Brain injury	1 (1.6%)
Brain malformation	1 (1.6%)

^*Specific point mutations, eponymous disorders, etc.

^†Children with significant/global developmental delay and clinical features of NI (for example, seizures) but without an underlying diagnosis listed in the chart

Performance of NI Diagnosis Codes

In hospitalized patients, the NI codes identified children with NI documented by health record with a specificity of 88.1% (95% CI: 84.7, 91), sensitivity of 79.4% (95% CI: 67.3, 88.5), NPV 96.7% (95% CI: 94.8, 98), and PPV 49% (95% CI: 42, 56.1). Lower NI prevalence scenarios demonstrated robustness of NPV, but decreasing NI prevalence from 5% to 2.5% decreased PPV from 26% (95% CI: 20.9-31.8) to 14.6% (95% CI: 11.4-18.5). (Supplemental Table 1) The Cohen’s Kappa for code list and expert designation showed moderate agreement (0.53, 95% CI: 0.43-0.64).

Effect of Higher and Lower Thresholds for Functional Impairment

When only children with severe functional impairments were considered to have NI (higher threshold), 52 children (10.4%) met criteria and the diagnosis codes had an NPV of 98.2% and PPV of 44.1%. When the 44 children with mild-to-moderate functional impairments were also categorized as NI (lower threshold), there were 96 children (19.2%) with NI, and the NI diagnosis codes had an NPV of 95.7% and a PPV of 77.5%. (Supplemental Table 2)

Evaluation of Misclassified Charts

Thirteen children had clinical documentation of NI but did not have NI discharge diagnosis codes (false negatives). On assessment of all ICD-10 codes for false negative cases, we did not find any diagnosis codes present in the children’s records that should have been on the NI diagnosis code list (that is, there were no false negative cases caused by NI diagnosis code list error). Reasons for false negative status included: a rare or unknown underlying diagnosis (n=6, 46.2%), a description of “significant” or “global” developmental delay with clinical features consistent with NI without a specified underlying diagnosis (n=4), and a reason for admission unrelated to NI with omission of the known NI diagnosis from the discharge diagnoses (n=3, 30.8%). A majority of cases with false negative status had one hospitalization over 2 years (n=10, 76.9%). Among the false positive cases (n=52), misclassification was attributed to the following: not meeting threshold for functional impairment (n=24, 46.2%), no signs of functional impairment (n=15, 28.8%), technology not related to NI (n=7, 13.5%), inadequate information to assess functional impairment (n=4, 7.7%) and misclassification by expert reviewers (n=2, 3.8%).

Linking Clinical NI Status to Administrative Data Characteristics

Children with clinically documented NI are more complex than those with milder functional impairments (“less severe impairment”) and those without NI. Children with clinical NI are more likely than children without NI to have multiple NI diagnoses falling into different categories, more admissions over the 2 years prior to the index admission, more non-NI complex chronic conditions, and more types of technology dependence. Children with degrees of functional impairment near threshold on chart review fell in between children with NI and children without NI on health administrative data markers of complexity. There was a higher proportion of children ≤1 year among the “less severe impairment” group (45.5%) compared to their overall prevalence (29.6%) because the smaller gap between developmental and chronologic ages made functional limitation more difficult to assess. Noting that children with NI were about 8 times more likely to have a feeding tube than children without NI (34.9% compared to 4.3%), we tested the effectiveness of a more restrictive method to identify NI that required both the neurologic diagnosis and a feeding tube diagnosis code. The test statistics of the more restrictive method compared to chart-review NI were: specificity of 98.4% (95% CI: 96.7-99.3), sensitivity of 28.6% (19.4-41.3), PPV of 61.5% (95% CI: 47-74.2), and NPV of 91.5% (95% CI: 89.9-92.9).

Discussion

The high-intensity NI diagnosis codes accurately identified most children with clinical NI on chart review, but also identified some children without functional impairment. Compared with health record documentation, the high-intensity NI diagnosis codes¹² were moderately discriminative in identification of children with NI and documented functional impairment (sensitivity of 79.4% and specificity of 88.1%). In a children’s hospital setting (NI prevalence 12.6%), the diagnosis codes appropriately excluded almost all children without NI (NPV 96.7%), but was overly inclusive (PPV 49%). Using the diagnostic codes in a lower prevalence population (for example, a community hospital with 2.5% NI), would lead to a high false positive rate (PPV 14.6%). Evaluation of misclassified charts demonstrated that the main NI diagnosis code performance issue was inclusion of children with mild or no functional impairment. Of evaluated health administrative markers of functional impairment, presence of a feeding tube was markedly more common among children with NI. As expected, incorporating codes for a feeding tube into the NI identification strategy increased specificity and lowered sensitivity, but improved the PPV only modestly (61.5%).

The performance of the NI diagnosis codes is likely adequate for health administrative studies describing patterns of healthcare utilization over time among children with NI^{4, 17, 25}—the original reason the code list was developed⁴—as the goal is to describe general trends instead of generating precise estimates. However, the population heterogeneity indicated by the low PPV is more problematic in clinical studies. With the growth of complex care as a field,¹⁰ the use of the NI construct in clinical studies is increasing.^18–20 This transition necessitates formal assessment of case definition validity because of the risk that misclassification bias—inaccurate assignment of case/non-case status—can lead to incorrect clinical conclusions.²⁶ There is no generally accepted level of performance—across 7 validation studies of health administrative data case definitions and algorithms^{22, 27–32} sensitivity ranged from 65%²⁷ to 97%³² and specificity from 75%²⁸ to 99%.³⁰ Comparatively, the high-intensity NI diagnosis codes we tested—with sensitivity of 79% and specificity 88%—had mid-range performance. However, the acceptability of that performance varies by study depending on the specific research question, whether misclassification would likely bias results toward or against the null, and the potential for biased results to cause harm.

In addition to estimating performance, validation studies can also highlight the clarity—of lack thereof—in case definitions for a given condition.³³ We identified a large “grey zone” in our chart review—children with a relevant diagnosis and moderate functional impairments or future risk of severe impairments. Classification in these cases was challenging, as indicated by the modest agreement between clinicians for the 9% “unsure” cases (Cohen’s kappa of 0.57). These discrepancies arose because the definition of NI—underlying neurologic diagnosis causing functional impairment⁴—is open to varying interpretations. Therefore, a major limitation of this study is that our test statistics incorporate more uncertainty than the “gold standard” designation of the chart review would suggest. Challenges in operationalizing the NI construct are common. A recent systematic review highlighted the varying definitions of severe NI in the literature, spanning types of impairment, as well as characteristics of the underlying disorder and clinical outcomes¹¹ reinforcing that the threshold for classification of NI is imprecise and often specific to the study context. The same group generated a consensus definition of severe NI—specifying that it encompasses multiple domains of impairment and causes complex medical needs and high care needs³⁴—but the new definition still requires clinical judgement for case classification. Our findings further emphasize the difficulties in assessment of functional impairment. While clinical scales like the Gross Motor Function Classification System (GMFCS)³⁵ demonstrate that stratification by functional level is possible, translation of their constructs into discharge diagnosis codes has not yet been done.

Our findings must be interpreted in the context of additional limitations. First, this study was from a single Canadian freestanding children’s hospital, limiting generalizability. Further validation studies, however, are unlikely to substantially improve estimates: while coding and charting practices may vary between institutions, the uncertainty from clinical classification likely far outweighs inter-institutional variation. Second, we noted suboptimal documentation of functional status markers within acute care hospitalization charts. For example, we found reference to wheelchair use in 2 of 500 charts. Given the importance of function and its inclusion in critical conceptual definitions (for example, the World Health Organization’s International Classification of Function³⁶ and pediatric medical complexity¹⁰), documentation of functional status is potential target for quality improvement. Third, this study required translation of American ICD-10 to Canadian ICD-10 codes, which may have generated errors, though hand-review of misclassified charts did not identify missing codes on the translated list. Fourth, we anticipated adequate description of function in a single hospital record to allow classification, but the sparse records of intervention-focused admissions (for example, admission for surgery or chemotherapy) led to 4 false-negative cases (7.7%).

As harnessing “big data” for clinical research becomes more commonplace, validation studies are necessary to evaluate how closely the population identified by diagnosis codes matches the clinical population of interest. However, the design of validation studies testing conceptual definitions without a clear gold standard (for example, widely accepted diagnostic criteria) is not simple. We learned several valuable lessons in this study. First, we prioritized calculating prevalence of NI in our choice of sampling frame (random selection of all admissions). Limiting the sampling frame to first general pediatrics admission in 2019 would have increased the number of positive cases, minimized charts without adequate documentation of function, and avoided duplicate sampling of children with multiple admissions. Second, chart reviews assessing conditions without a clear gold standard benefit from involvement of multiple clinicians to refine case definitions during the training phase. Determining the appropriate number of charts to be reviewed by multiple clinicians requires finding the right balance between potential for agreement and efficiency, as increasing numbers of charts with multiple reviewers improves the likelihood of high agreement but requires more investment of time. We would recommend finalizing the secondary review plan after the training phase to find an appropriate balance.

Conclusions

The high-intensity NI diagnosis codes effectively identify the population of children at risk for NI, as demonstrated by the consistently high NPV, but also include children without NI-related functional impairment (lower PPV). Studies using the NI diagnosis codes should describe the risk of misclassification bias and consider use of design or statistical strategies to minimize it.^{37, 38} This study also highlights the importance of validation studies to clinically evaluate the population accrued using diagnosis codes. Even for widely used diagnosis code lists with strong face validity, clinical evaluation may demonstrate meaningful gaps between the expected and observed population that should be addressed in future studies relying on the diagnosis codes for population ascertainment.

Table 3.

Cohort Characteristics derived from Administrative Data Stratified by Chart-Review-Verified Neurologic Impairment Status.

	No NI (N=404)	Less severe impairment N=44 (11 with NI and 33 without NI)	NI (N=52)
NI discharge code in 2 years of hospitalization data, n (%)
No NI code	381 (94.3%)	10 (22.7%)	7 (13.5%)
NI code	23 (5.7%)	34 (77.3%)	45 (86.5%)
Age category, n (%)
≤1 year	115 (28.5%)	20 (45.5%)	13 (25.0%)
2-5 years	72 (17.8%)	4 (9.1%)	12 (23.1%)
6-11 years	100 (24.8%)	8 (18.2%)	16 (30.8%)
≥12 years	117 (29.0%)	12 (27.3%)	11 (21.2%)
Sex, n (%)
Female	187 (46.3%)	14 (31.8%)	26 (50.0%)
Male	217 (53.7%)	30 (68.2%)	26 (50.0%)
Number of NI diagnosis categories, n (%)
None	381 (94.3%)	10 (22.7%)	7 (13.5%)
1 category	20 (5.0%)	26 (59.1%)	22 (42.3%)
2 categories	3 (0.7%)	6 (13.6%)	16 (30.8%)
3 categories	0 (0%)	2 (4.5%)	7 (13.5%)
Type of NI diagnostic category, n (%) *
Epilepsy, n (%)	11 (2.7%)	7 (15.9%)	22 (42.3%)
Anatomic, n (%)	6 (1.5%)	10 (22.7%)	14 (26.9%)
Peripheral, n (%)	3 (0.7%)	4 (9.1%)	4 (7.7%)
Static, n (%)	3 (0.7%)	10 (22.7%)	15 (28.8%)
Stroke/hemorrhage, n (%)	2 (0.5%)	4 (9.1%)	3 (5.8%)
Genetic, n (%)	1 (0.2%)	6 (13.6%)	13 (25%)
Metabolic, n (%)	0 (0%)	3 (6.8%)	1 (1.9%)
Progressive/movement, n (%)	0 (0%)	0 (0%)	3 (5.8%)
Admissions over 2-year period, n (%)
1 admission	254 (62.9%)	21 (47.7%)	16 (30.8%)
2-3 admissions	72 (17.8%)	12 (27.3%)	18 (34.6%)
4-9 admissions	61 (15.1%)	7 (15.9%)	8 (15.4%)
≥10 admissions	17 (4.2%)	4 (9.1%)	10 (19.2%)
Number of non-NI complex chronic conditions, n (%)
None	204 (50.5%)	9 (20.5%)	8 (15.4%)
1 type	133 (32.9%)	17 (38.6%)	19 (36.5%)
2 types	49 (12.1%)	12 (27.3%)	7 (13.5%)
≥3 types	18 (4.5%)	6 (13.6%)	18 (34.6%)
Technology dependence, n (%)
None	363 (89.9%)	27 (61.4%)	29 (55.8%)
1 type	17 (4.2%)	9 (20.5%)	11 (21.2%)
2 types	18 (4.5%)	6 (13.6%)	7 (13.5%)
≥3 types	6 (1.5%)	2 (4.5%)	5 (9.6%)

^*One child may belong to multiple categories;

^**Non-NI neuromuscular CCC category includes peripheral nervous system diagnoses (for example, anterior horn disorder) and isolated cognitive impairment;

^***Within the technology dependence “Other” category, 22/28 (79%) of children have the same diagnosis code for an unspecified “external stoma.”

What’s New.

Children with neurologic impairment (NI) are often identified in health administrative data using discharge diagnosis codes. This study reviewed charts of hospitalized children to determine how well the NI diagnosis codes identify children with the functional impairment expected in clinical NI.