Pathophysiological Features in Electronic Medical Records Sustain Model Performance under Temporal Dataset Shift.

Access to real-world data streams like electronic medical records (EMRs) has accelerated the development of supervised machine learning (ML) models for clinical applications. However, few studies investigate the differential impact of particular features in the EMR on model performance under temporal dataset shift. To explain how features in the EMR impact models over time, this study aggregates features into feature groups by their source (e.g. medication orders, diagnosis codes and lab results) and feature categories based on their reflection of patient pathophysiology or healthcare processes. We adapt Shapley values to explain feature groups' and feature categories' marginal contribution to initial and sustained model performance. We investigate three standard clinical prediction tasks and find that while feature contributions to initial performance differ across tasks, pathophysiological features help mitigate temporal discrimination deterioration. These results provide interpretable insights on how specific feature groups contribute to model performance and robustness to temporal dataset shift.

Prevalence of pediatric lower urinary tract symptoms in a national claims database of privately insured patients, 2003-2014.

BACKGROUND: We conducted this study to estimate the prevalence of pediatric lower urinary tract symptoms (pLUTS) in a US privately insured pediatric population who are 6-20 years old by age, sex, race/ethnicity from 2003-2014. This has not been previously described in the literature. METHODS: We retrospectively reviewed Optum's de-identified Clinformatics® Data Mart Database between 2003-2014. A pLUTS patient was defined by the presence of ≥ 1 pLUTS-related ICD-9 diagnosis code between the age of 6-20 years. Neurogenic bladder, renal transplant and structural urologic disease diagnoses were excluded. Prevalence by year was calculated as a proportion of pLUTS patients among the total population at risk. Variables reviewed included age, sex, race, geographic region, household factors and clinical comorbidities including attention-deficit/hyperactivity disorder (ADHD), constipation, and sleep apnea. Point of service (POS) was calculated as a proportion of pLUTS-related claims associated with a POS compared to the total claims at all POS in the time period. RESULTS: We identified 282,427 unique patients with ≥ 1 claim for pLUTS between the ages of 6-20 years from 2003 to 2014. Average prevalence during this period was 0.92%, increasing from 0.63% in 2003 to 1.13% in 2014. The median age group of patients was 6-10 years. More patients were female (59.80%), white (65.97%), between 6 and 10 years old (52.18%) and resided in the Southern US (44.97%). Within a single household, 81.71% reported ≤ 2 children, and 65.53% reported ≥ 3 adults. 16.88% had a diagnosis of ADHD, 19.49% had a diagnosis of constipation and 3.04% had a diagnosis of sleep apnea. 75% of pLUTS-related claims were recorded in an outpatient setting. CONCLUSIONS: Families consistently seek medical care in the outpatient setting for pLUTS. The demographic and clinical characteristics of our cohort reflect prior literature. Future studies can help define temporal relationships between household factors and onset of disease as well as characterize pLUTS-related healthcare resource utilization. Additional work is required in publicly insured populations.

Prevalence of pediatric lower urinary tract symptoms in a US population seeking medical care, 2003- 2014.

Background: We conducted this study to estimate the prevalence of pediatric lower urinary tract symptoms (pLUTS) in a US privately-insured pediatric population who are 18 years of age or older by age, sex, race/ethnicity from 2003-2014. This has not been previously described in the literature. Methods: We retrospectively reviewed Optum's de-identifed Clinformatics® Data Mart Database database between 2003-2014. A pLUTS patient was defined by the presence of ≥ 1 pLUTS-related ICD-9 diagnosis code between the age of 6-20 years. Neurogenic bladder, renal transplant and structural urologic disease diagnoses were excluded. Prevalence by year was calculated as a proportion of pLUTS patients among the total population at risk. Variables reviewed included age, sex, race, geographic region, household factors and clinical comorbidities including attention-deficit/hyperactivity disorder (ADHD), constipation, and sleep apnea. Point of service (POS) was calculated as a proportion of pLUTS-related claims associated with a POS compared to the total claims at all POS in the time period. Results: We identified 282,427 unique patients with ≥ 1 claim for pLUTS between the ages of 6-20 years from 2003-2014. Average prevalence during this period was 0.92%, increasing from 0.63% in 2003 to 1.13% in 2014. Mean age was 12.15 years. More patients were female (59.80%), white (65.97%), between 6-10 years old (52.18%) and resided in the Southern US (44.97%). Within a single household, 81.71% reported ≤ 2 children, and 65.53% reported ≥ 3 adults. 16.88% had a diagnosis of ADHD, 19.49% had a diagnosis of constipation and 3.04% had a diagnosis of sleep apnea. 75% of pLUTS-related claims were recorded in an outpatient setting. Conclusions: Families consistently seek medical care in the outpatient setting for pLUTS. The demographic and clinical characteristics of our cohort reflect prior literature. Future studies can help define temporal relationships between household factors and onset of disease as well as characterize pLUTS-related healthcare resource utilization. Additional work is required in publicly-insured populations.