Joint effects of advancing age and number of potentially inappropriate medication classes on risk of falls in Medicare enrollees

Table 4 Logistic regressions of fall events on FRD/PIM measures and demographic characteristics^a

	Approach to Measurement of Effect of FRD/PIMs
	Model 1 Individual Medications^b			Model 2 Counts of FRD/PIM Classes			Model 3 NNH-Based Risk Categorizations
N of cases included in model	1,678,037			1,678,037			1,678,037
Model chi-square	108713.4***			105033.2***			88936.9***
C-statistic (area under ROC curve)	0.774			0.768			0.748
Nagelkerke R square	0.163			0.158			0.134
	AOR	CI-L	CI-U	AOR	CI-L	CI-U	AOR	CI-L	CI-U
Age 65–74	REF	REF	REF	REF	REF	REF
Age 75–84	1.751	1.723	1.778	1.738	1.711	1.765
Age 85–89	2.830	2.774	2.887	2.802	2.747	2.859
Age 90–94	3.771	3.674	3.870	3.739	3.644	3.837
Age 95–99	4.812	4.579	5.056	4.760	4.531	5.000
Female	1.339	1.321	1.358	1.384	1.365	1.403	1.402	1.383	1.421
Facility use for fracture^c	1.194	1.159	1.231	1.271	1.233	1.310	1.368	1.328	1.410
Anticholinergic	1.073	1.055	1.092
Benzodiazepine	1.053	1.035	1.071
Antidepressant (any type)	1.336	1.314	1.358
Loop diuretic	1.174	1.154	1.194
Antimuscarinic	1.160	1.130	1.192
Antipsychotic, first-generation	1.219	1.115	1.332
Neuropathic pain reliever	1.092	1.070	1.115
Opioid pain reliever	1.462	1.439	1.486
No FRD/PIM classes (NNH reference category)				REF	REF	REF	REF	REF	REF
1 FRD/PIM class				1.021	0.998	1.044
2 FRD/PIM classes				1.128	1.102	1.154
3 FRD/PIM classes				1.243	1.213	1.273
4 FRD/PIM classes				1.357	1.323	1.393
5 or more FRD/PIM classes				1.579	1.540	1.619
NNH 200–333							0.850	0.829	0.871
NNH 83–91							1.127	1.098	1.156
NNH 17–48							1.427	1.398	1.456
NNH < 15							1.983	1.934	2.032

^aAll models adjusted for the following diagnoses: Atrial fibrillation, cancer, dementia, depression, diabetes, dizziness/syncope, gait disorder, hepatic impairment, hypertension, neuropathy, orthostatic hypotension, Parkinson’s disease, renal impairment, and substance abuse. Impaired vision was removed in a backwards stepwise logistic regression analysis (P in = 0.001, P out = 0.005) of fall events on diagnosis controlling for age group, sex, and use of a facility for fracture, where diagnoses were measured from 12 months prior through 7 days after the event or proxy date, and facility usage was measured from 12 months prior through 31 days prior to the event or proxy date. ^bBackwards stepwise analysis of all FRD/PIM classes with unadjusted ORs of 1.50 or more. ^cUse of a facility for fracture, measured from 12 months prior through 31 days prior to the event or proxy date. AOR adjusted odds ratio, CI-L and CI-U lower and upper limits, respectively, of the 95% confidence interval, FRD fall risk drug, NNH number needed to harm, PIM potentially inappropriate medication, REF reference group, ROC receiver operating characteristics

ISSN: 1471-2318