We received claims data for 2014 from PFK for 258,399 pediatric patients. Demographic features and summary statistics for the study population are described in Table 3.
The mean (±SD) age of patients, using oldest age in 2014, was 8.7±5.1. The population was 51.1% male and 48.9% female. Overall, 58,762 (22.7%) patients were category 1 (n=11,672 as 1-A and n=47,090 as 1-B). Category 2 included 60,307 (23.3%) patients, while the remaining patients (n=139,330; 53.9%) were category 3. The claims data encompassed 6,962 IP hospitalizations (admissions from the ER), 230,384 OP-ER visits, and 792,129 OP office visits.
Overall, the mean 2014 annual cost per 1-A child ($4,709) was nearly six times greater than category 3 children ($786). Annual costs for 1-A children were significantly higher than 1-B children, which were significantly higher than category 2, while category 3 had the lowest costs (Table 4A). We observed this significant, stepwise decrease in paid amounts and number of visits across the four categories in most of our analyses (Table 4).
IP costs were calculated by summing paid claims from facility claims and corresponding professional claims data (Table 4B). IP claims include scheduled admissions as well as ER visits resulting in hospitalization. Here, ER may also refer to urgent care claims, as they are labeled the same in PFK claims database.
The proportion of children who acquired IP costs in 2014 significantly decreased from category 1 to category 3; in fact, 6.1% of 1-A children accumulated IP charges in 2014, compared to 3.8% (category 1-B), 2.8% (category 2), and 1.1% (category 3). While >97% of our study population did not experience IP admissions, the odds ratio (OR) of category 1 (A&B) genetic patients experiencing ≥1 admission was OR=4.12 (95% CI:3.86-4.39, p<0.0001), compared to category 3 non-genetic patients (Figure 2).
Specifically, children in 1-A averaged 1.5 IP visits in 2014 versus 1.2 visits (p<0.0001) for all other children in our study population. 1-A patients had average paid claims of $17,457 per admission, which was nearly 28% higher than other genetic patients in 1-B ($13,668 per admission); 1-A and 1-B were both significantly more costly for IP admissions than categories 2 ($8746) and 3 ($7595). Average LOS for IP admissions in 1-A (5.1 days; CI:4.3-5.8) was significantly higher than 1-B (4.1 days; CI:3.7-4.5), category 2 (3.5 days; CI:3.3-3.7), and category 3 (3.8 days; CI:3.2-4.4).
OP office costs were calculated by summing paid claims in the data flagged as “office visit.” We included scheduled office visits except for dental, vision, and mental health. Additionally, we included all costs (e.g., therapies, diagnostic testing) associated with the visit. We identified OP visits flagged as “ER” and included all associated charges with the ER visit and separate our analyses into OP office and OP-ER visits, where OP-ER visits indicate ER visits not resulting in admission (Table 4C).
Annual mean costs for OP office visits per child were highest in 1-A ($560) and consistently decreased across all categories, with lowest costs in category 3 children ($181). The proportion of children who accumulated OP office costs in category 1 was ~88% (both 1-A and 1-B), significantly higher than the proportion of category 2 (~86%) and category 3 (~71%) who acquired OP office costs. Children with single gene disease and chromosomal disorders (1-A) averaged 5.3 visits annually, significantly higher than the average of our entire study population at 3.9 visits. On a per visit cost basis, OP office visit paid claims were highest in 1-A at $120, a price significantly higher than other categories.
We observed similar trends in OP-ER data as we did in OP office data (Table 4C). For example, category 1-A patients accounted for higher annual costs, higher number of annual visits, and higher cost per visit, compared to all others. Specifically, 54.5% of 1-A patients had at least one OP-ER visit, compared to 43.7% of our entire study population. The annual OP-ER paid amounts per child was highest in 1-A ($238), identical in 1-B and 2 ($197), and lowest in category 3 ($108).
Average Rx paid amounts per child in 2014 was highest for 1-A patients at $1370, significantly higher than the average for all children in our study population ($363), and significantly more than any other category (Table 4D). Rx claims data only includes paid amounts for OP medications.
In Table 4E, we display “other costs” to include paid amounts from our data not categorized as IP, OP office, OP-ER, or Rx costs. These costs may include home health visits, OP surgeries, dental, or vision care, to name a few. Although these costs come from varied categories of charge, it is of interest to note we observed significant differences between the categories of genetic patients, with the highest paid amounts in 1-A patients.
Children Without Costs
Overall, 8.2% (n=21,303) of our population did not have any claims filed for 2014; that is, their 2014 costs were zero, even though they had continuous Medicaid coverage for the entire year. We did, however, categorize these patients based on ICD-9 diagnoses from previous years and included them in some analyses as described in our study design methods section. Of the 21,303 children without any filed claims, the majority (n= 16,886; 79.3%) were category 3, or non-genetic patients.
Most Common and Highest Cost ICD-9 Diagnoses
We investigated which specific ICD-9 codes per category affected the most patients (i.e., most common, Table 5) and which diagnoses were associated with the highest costs per child (Table 6). Although not the focus of this study, it would be of interest to know if there are specific genetic conditions significantly driving costs, which may help ACOs to initially target these patients for disease management and care coordination. The most common genetic diagnoses were hereditary hemolytic anemias, for example sickle cell disease (1-A) and symptoms concerning nutrition and development, for example failure to thrive (1-B), which often have significant developmental and genetic factors. The most common category 2 condition was asthma, while the most common category 3 ICD-9 code was V06 (i.e., childhood vaccinations). As expected, these most common codes were associated with highest costs in 2014 overall, given the high number of patients attributed to each code. For example, 1,227 patients in our study population received a diagnosis of cystic fibrosis (1-A) and were associated with >$19M in paid claims. When normalized to “cost per child,” the costliest genetic disorders were aplastic anemias, for example Fanconi anemia (1-A), and disorders of parathyroid gland (1-B). Among category 2, malignant neoplasms, particularly of bone and articular cartilage, were associated with highest costs on a per child basis. In fact, malignant neoplasms in general accounted for four of the top five most expensive category 2 diagnoses. Among non-genetic diagnoses in category 3, phlebitis and thrombophlebitis carried the highest costs per child.
Manual Review of Electronic Health Records
For our study population, we aimed to measure the classification accuracy of our genetic categorization method (Figure 3). We sampled 100 patients to confirm if the genetic categorization based on claims data (i.e., predicted categorization) could be validated in their electronic health chart (i.e., true categorization). First, we determined four cost quantiles for all patients and then randomly selected 25 patients from each quantile. Next, for all 100 patients, we manually searched electronic health charts to categorize patients based on the categories described in Table 1. We observed agreement between claims data and electronic health records in 81/100 patients. In other words, we confirmed our claims-based genetic categorization (1-A, 1-B, 2, or 3) in 81% of our study population using health chart records.