Using machine learning to identify patterns of lifetime health problems in decedents with autism spectrum disorder.
Lauren Bishop-Fitzpatrick, Arezoo Movaghar, Jan S Greenberg and 4 others
PMID 29734508WHAT IT FOUND
People with autism who died had more lifetime health problems, including epilepsy, long-term medication use and side effects, motor problems, choking, urinary, digestive and ear issues, but fewer records for alcohol abuse, hypertension and metastatic cancer.
These are patterns, not treatment effects.
Key findings
01The model distinguished decedents with ASD from matched controls with sensitivity 0.75, specificity 0.94, and accuracy 0.93.
02Distinguishing lifetime code patterns included higher rates of epilepsy, long-term medication use and side effects, developmental problems, ear problems, non-specific lab tests and encounters, urinary problems, digestive problems, motor problems, and choking.
03After controlling for age and sex, decedents with ASD differed from controls on 12 of 28 standard comorbidity categories, with increased risk for coagulopathy, congestive heart failure, deficiency anemia, fluid and electrolyte disorders, hypothyroidism, paralysis, valvular disease, weight loss, and other neurological disorders, and decreased risk for alcohol abuse, hypertension, and metastatic cancer.
STILL TO COME
How it was doneWhat they foundWhat it means for PTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
It studied people who had died, so it cannot describe health problems in living children, adolescents or adults with ASD. The final matched ASD group was small: 91 decedents with ASD. Most ASD decedents and controls were aged 40 or older at death, so the findings mainly concern midlife and old age. Health problems were captured from electronic diagnostic codes, which may be incomplete or inaccurate, and mental disorder codes were excluded. The study cannot show why these code patterns occur; lower risks for some conditions could reflect less screening or diagnosis. Matching was imperfect because sex and birth year distributions differed, and not all ASD decedents could be matched at 1:68.
Declared interests
No funding or conflicts declaration appears in the supplied article text. The publication types list NIH extramural research support.
The easy way to misread this
Do not read the model accuracy or the increased and decreased comorbidity risks as proof that autism causes these health problems or that therapy changes mortality. The study used retrospective electronic records from decedents, excluded mental disorder codes, and cannot separate true lower risk from less screening or diagnosis.