Mental illness and intensification of diabetes medications: an observational cohort study

Table 1 Construction of study cohort

Sample size n	Criterion	MHC Yes	MHC No
		n(%)	n(%)
440,953	Diabetic veteran VA outpatient treated with oral antiglycemic (during the 6 months prior to the OI), who was alive at start of OI and for whom MHC status can be determined	80,798	360,155
	↓
427,335	VA outpatient use in OI*	80,745	346,590
		100%	96%
	↓
426,605	Non-institutionalized in OI^†	80,344	346,261
		100%	100%
	↓
426,454	No data problems^‡	80,323	346,131
		100%	100%
	↓
366,066	No conditions likely to alter goals of care^§	67,099	298,967
		84%	86%
	↓
315,063	Patient’s home facility submitted usable HbA1c lab data to central data repository	57,309	257,754
		85%	86%
	↓
269,692	Had at least one HbA1c test completed during the OI	51,582	218,110
		90%	85%
	↓
	Poor glycemic control: Had at least one HbA1c ≥8 during OI eligible to serve as an index HbA1c:	20,803	71,572
		36%	28%
	(1) no antiglycemic intensification in the 3 months prior to the HbA1c test, AND	18,802	65,458
		90%	91%
	(2) no hospitalization in the 3 months prior to the HbA1c test, AND	18,197	64,670
		97%	99%
	(3) not occurring on a hospital admission day, AND	18,180	64,629
	(3) not occurring on a hospital admission day, AND	100%	100%
58,364	(4) no insulin prescribed in the 6 months prior to the HbA1c test	11,581	46,783
58,364		64%	72%
	↓
52,526	Final analytic cohort: Testing data set (90% random sample)	10,422	42,104
		90%	90%

Legend: This table shows sample size at each step of cohort construction, overall and for those with/ without Mental Health Conditions; % reported in a row refers to % of prior row remaining after applying the inclusion criterion listed in current row.
*Utilization refers to VA face-to-face outpatient care of any type; telephone, laboratory or radiology encounters did not qualify as face-to-face. Note that percentages in each row use the number in the prior row as denominator.
^†Did not spend more than half the year in a VA inpatient or long-term care setting.
^‡Data quality issues such that unique identifier (scrambled social security number), date of birth or vital status was indeterminate.
^§End-stage renal disease, end-stage liver disease, cancer, stroke or dementia diagnosis occurring in the two years prior to the observation interval.
Abbreviations: OI, Observation Interval (April 1, 2003 through March 31, 2004); MHC, Mental Health Condition; HbA1c, Hemoglobin A1c; VA, Veterans Health Administration.

ISSN: 1472-6963