The Association between Nursing Home IT Maturity and UTI among Long-Term Residents

Abstract

Urinary tract infections (UTIs) are the most common infections among nursing home (NH) residents. Antibiotics are often misused when a UTI is suspected. Using sophisticated information technology (IT) could help in the appropriate UTI prevention, diagnosis and treatment. This repeated cross-sectional study explored relationships between IT maturity and UTI prevalence among long-stay NH residents. Data were from 1) four annual surveys 2013–2017 measuring IT maturity in a random sample of Medicare-certified NHs, 2) Minimum Data Set assessments for resident characteristics, and 3) Certification and Survey Provider Enhanced Reporting data for facility characteristics. In multivariate regressions using NH fixed effects, controlling for resident and NH characteristics, Administrative IT maturity in NHs was associated with decreased odds of UTI (AOR: 0.906, 95% CI: 0.843, 0.973). These results were robust in all sensitivity analyses. Using IT to relieve administrative burden may decrease UTIs.

Background

Urinary tract infections (UTIs) are the most common infections among nursing home (NH) residents (Herzig et al., 2017), often resulting in preventable hospitalization and other negative outcomes (Clinic, 2018). It was previously reported that a majority of antibiotic use for UTI was inappropriate in NH (Pulia et al., 2018), which may be due to high prevalence of asymptomatic bacteriuria among residents, for which antibiotics are not warranted (Nace, Drinka, & Crnich, 2014). Antibiotic overuse is a public health crisis because it contributes to the rise of antimicrobial resistance. Improving UTI prevention and management in NHs is included in key antibiotic stewardship and infection control and prevention initiatives (Morrill, Caffrey, Jump, Dosa, & LaPlante, 2016; Palms et al., 2019).

Information technology (IT) systems that facilitate tracing high-risk individuals, identifying antibiotic clinical indications, enhancing culture results, and offering decision support could help staff identify UTIs and proper interventions earlier (Felix, Dayama, Morris, Pradhan, & Bradway, 2021). However, disparities in adoption and use of NH IT systems caused by lack of investment in IT infrastructure, preparation, and training could place vulnerable residents at greater risk of UTI (Ko, Wagner, & Spetz, 2018). While NH IT adoption is increasing nationally, disparities in use are widening as some facilities are abandoning IT systems due to perceived usability and efficiency problems (G.L. Alexander, Madsen, & Newton, 2017). Jones et al. found that 14% of NHs in a statewide study used sophisticated software packages to support infection control activities and less than 20% had system integration between radiology, laboratory, and electronic medical records (Jones, Samore, Carter, & Rubin, 2012). Conceptually, NH IT maturity (i.e., technological capability, extent to which systems are used, and degree to which different systems are integrated) could impact the NH’s ability to prevent and manage UTIs through improved integration and communication of data from multiple sources such as pharmacy, laboratory, radiology records and other electronic medical records (Davidson & Jump, 2020). In prior work, we found significant inverse correlations (r = −.19, p < .01) with NH UTI prevalence and trends in IT adoption in resident care over three years (Gregory L. Alexander, Madsen, Deroche, Alexander, & Miller, 2020). However, to our knowledge, no IT maturity measures have been linked to health outcomes at the resident-assessment level. The aim of our study was to explore the relationship between specific aspects of IT maturity and odds of UTI among long-stay NH residents. This information could shape utilization of IT in NHs and thus may improve care quality.

Methods

This repeated measures study was approved by the Institutional Review Boards of the University of Missouri, Columbia University and RAND Corporation. The authors used the Quality Health Outcomes Model to inform the design of this study (See Figure) (Mitchell, Ferketich, & Jennings, 1998).

Figure.

The Conceptual Model framing this analysis, based on the Quality Health Outcomes Model (Mitchell et al., 1998).

Data Sources

We used three data sources: NH IT maturity from four annual national surveys, resident health and demographic information from the Minimum Data Set (MDS) 3.0, and NH characteristics from Certification And Survey Provider Enhanced Reports (CASPER). The time horizon was 2013–2018, representing years in which all data sources were available to us.

IT maturity items were collected from NH Administrators or appointed persons who had oversight of IT systems, survey methods were described previously (G. L. Alexander & Madsen, 2009; Gregory L. Alexander & Madsen, 2018; G. L. Alexander, Madsen, & Wakefield, 2010; G. L. Alexander et al., 2017). We randomly sampled Center for Medicare and Medicaid (CMS)-certified NHs across the nation, excluding those in U.S. territories, Veterans Administration government facilities, facilities designated as special focus NHs, i.e., just over 1% of U.S. facilities had a special focus with quality problems and thus subject to non-generalizable interventions. A total of 817 NHs completed at least one year of the survey during 2013–2017. Response rates were 45% - 56% each year (G. L. Alexander, Madsen, Deroche, Alexander, & Miller, 2019). The survey measured NH IT maturity along three healthcare domains: resident care, clinical support, and administrative activities. Within each of the three IT maturity dimensions IT capabilities, extent of use, and degree of integration was defined. Therefore, this 3 × 3 (domains/dimensions) model has nine subscales each with a minimum of 0 and maximum of 100. These were the independent variables of interest. For sensitivity analyses (discussed below), we also used components of the domains/dimension, which were coded into zeros representing no IT maturity, and one or greater representing higher levels of IT maturity. The survey had good reliability and validity measures; the Cronbach alpha values for the three healthcare domains of resident care, clinical support, and administrative activities are 0.87–0.88, 0.86–0.91, and 0.70–0.80, respectively (G. L. Alexander et al., 2017).

We paired survey data by NH provider number and survey year with MDS, which includes detailed health assessments of NH residents (Bergman-Evans, 2020). We included MDS assessments conducted within 90 days of IT maturity survey completion, which was important to capture IT measures nearest to the time the MDS assessments were completed.

Inclusion & Exclusion Criteria

We included 816 NHs that completed the survey with at least one regular (i.e., quarterly and annual assessments), MDS assessment in the survey window per survey-year, which are not triggered by specific health-related events and thus reduce bias in our analyses (Bergman-Evans, 2020). We removed individuals under the age of 65 at the time of assessment (13% of residents) as their UTI risk differs from the majority of residents (Castle, Engberg, Wagner, & Handler, 2017).

Outcome

The outcome was a binary indicator of UTI, which is recorded on MDS assessments when: 1) the infection has been diagnosed by an advanced care provider in the last 60 days and 2) had “a direct relationship to the resident’s current functional status, cognitive status, mood or behavior, medical treatments, nurse monitoring, or risk of death” during the 30-day look-back period (Centers for Medicare & Medicaid Services, 2013). Further, the diagnosis must follow evidence-based criteria such as McGeer or Loeb that includes clinical symptoms (Centers for Medicare & Medicaid Services, 2013).

Covariates

From the MDS data, we described residents by characteristics thought to influence UTI presence and diagnosis, and those in prior literature (Castle et al., 2017). Thus, we examined resident demographics (age, race, gender) and clinical characteristics (behavioral problems, recent fall, depression, urinary incontinence, bowel incontinence, dementia, restraint use, pressure ulcers, anti-depressant medication, hypnotic medication, antipsychotic medication, and indwelling, intermittent, and external catheters). We defined dementia using Alzheimer’s Disease and/or Related Dementias diagnoses, included as two mutually exclusive variables, both with and without Advanced Cognitive Illness (ACI), as used previously (Cohen et al., 2021). Most MDS variables used in this study had very good to excellent reliability and validity (Saliba & Buchanan, 2008). We also controlled for seasonality (calendar month), time (calendar year) using the target date of the assessment, and the number of regular assessments for each unique resident per survey window accounting for the potential effects of some individuals being represented more often than others in the analysis.

Using CASPER data, we examined NH characteristics that are potential proxies for available resources and clinical processes that may affect UTI diagnoses: chain membership, % of Medicaid eligible residents, ownership type, size, staffing, and urbanicity.

Data Analysis

The unit of analysis was the resident assessment. We described sample characteristics, and the proportion of regular MDS assessments indicating UTI for each survey-year. We calculated means and standard deviations of various resident and NH characteristics with and without UTI. We compared these means (bivariate regressions with NH cluster standard errors) to determine the associations between UTIs and resident and NH characteristics.

We created a main model in which we regressed UTI on all IT maturity domain/dimensions, controlling for resident demographics, health status (including catheter use), NH characteristics, time, and seasonality. We included normalized domain/dimension measures to ease interpretation of final results. We excluded covariates from the regression that were fully dependent on one another (e.g., presence of an indwelling catheter and urinary incontinence), or had little or no variation with UTI. Total staffing, and percent of staffing provided by RNs were included. Because the NHs that had responded to the IT maturity survey differed from non-respondents with regard to ownership, location (G.L.; Alexander, Deroche, & Powell, 2020), we adopted post-stratification methods to reflect NH distributions with these characteristics for each state and included state-level survey weights to approximate a nationally representative sample. We computed robust standard errors clustered at the state-level to account for correlation among observations within state and multiple observations from the same person. We also included NH fixed effects to account for unobserved differences in individual NHs and to isolate the effects of IT maturity on UTI diagnoses within included NHs. We applied a 2-sided t-test with alpha = .05 and computed adjusted odds ratios (AOR) and 95% confidence intervals (CI). All analyses were conducted in Stata MP 15 (StataCorp, 2017).

Robustness Analyses

If a domain/dimension was associated in the main model, we ran an additional regression with the corresponding individual components rather than the composite measure. We also generated the main model without fixed effects, controlling for NH characteristics that do not vary over time (e.g., urbanicity). We created multivariate logistic regressions for each individual IT maturity domain/dimension without others included. Further, we created regressions similar to the main model, but stratified by cognitive status or catheter use to examine whether IT maturity might have different relationships with UTI among these various populations. Finally, we repeated our main model with a panel of NHs that had completed all four survey-years (N=186).

Results

Our sample included 816 NHs. These NHs had 219,730 regular NH resident assessments within 90 days of a survey, representing 80,322 unique long-term NH residents. Of these assessments, 4.1% recorded a UTI (See Table 1). Those with UTI had higher mean Activity of Daily Living (ADL) scores and were more likely to be female and have catheterization, incontinence, depression, or a fall. Those with UTI were less likely to have a dementia diagnosis (See Table 2).

Table 1.

Proportions of Minimum Data Set nursing home resident assessments^* with one-day antibiotic use, by survey year and conditions of interest (N=816 nursing homes)

Survey Year	# of NHs	# of Residents	# of Assessments	% of Assessments with UTI, by condition
				All	Sept.	All ADRD	ADRD, Without ACI	ADRD, With ACI
1	810	48,602	89,068	5.0%	20.4%	4.8%	5.1%	3.6%
2	452	26,816	49,460	4.0%	14.4%	3.8%	4.0%	3.2%
3	444	26,156	48,538	3.7%	18.1%	3.5%	3.6%	2.8%
4	325	18,953	32,664	2.6%	11.1%	2.4%	2.5%	2.0%
Overall	816	80,322	219,730	4.1%	16.1%	3.9%	4.2%	3.1%

Note: NH, Nursing Homes; Sept, Septicemia; ADRD, Alzheimer’s Disease or Related Dementia; ACI, Advanced Cognitive Impairment.

^*Represents quarterly and annual MDS assessments from the 816 U.S. Nursing Homes survey respondents, within 6 months of the 4 surveys (2013–2018).

Table 2.

Associations between IT maturity dimensions/domains and antibiotic use, determined by a logistic regression with NH fixed effects and state cluster robust standard errors. (N= 200,535)^*

Characteristics	All N= 219,730		UTI N= 9,027		No UTI N= 210,597
Resident
	Mean	SD	Mean	SD	Mean	SD	P
Age (Years)	83.86	8.82	83.90	8.48	83.86	8.84	.58
Activities of Daily Living Score: 0–28	16.43	6.70	17.64	5.72	16.38	6.73	< .001
	%	n	%	n	%	n	P
Female, %	73%	159,597	78%	7,013	72%	152,497	< .001
Race
White only	84%	181,272	87%	7,769	84%	173,403	< .001
African-American only	10%	21,186	7%	642	10%	20,543	< .001
Asian only	1%	2,207	1%	55	1%	2,152	< .001
Am.Ind./Al.Nat only	1%	1,249	1%	52	1%	1,195	.86
Hawaiian/Pac.Isl.	0%	170	0%	7	0%	163	.69
Hispanic only	4%	8,188	4%	340	4%	7,847	.88
Multi-racial	0%	1,066	0%	28	1%	1,037	< .001
Behavioral Problems	12%	27,186	14%	1,289	12%	25,881	.001
Catheter, Indwelling	4%	9,642	14%	1,268	4%	8,360	< .001
Catheter, Intermittent	0%	507	1%	89	0%	418	< .001
Catheter, External	0%	236	0%	14	0%	222	.60
Alzheimer’s Disease and Related Dementia (ADRD)	64%	141,395	62%	5,569	64%	135,775	.003
ADRD, with Advanced Cognitive Illness (ACI)	13%	29,269	10%	911	13%	28,345	< .001
ADRD, no ACI	51%	112,136	52%	4,659	51%	107,437	.21
	54%	119,228	58%	5,214	54%	113,950	< .001
	22%	48,132	27%	2,402	22%	45,693	< .001
Urinary Incontinence	82%	171,545	88%	6,895	81%	164,570	< .001
Bowel Incontinence	62%	134,513	69%	6,067	62%	128,391	< .001
Restraints Used	13%	27,814	11%	971	13%	26,833	0.349
Nursing Home
	Mean	SD	Mean	SD	Mean	SD	P
NH Size (# of Beds)	126.53	69.86	123.65	67.09	126.64	69.98	.13
% of staff at RN level	0.18	0.06	0.18	0.06	0.18	0.06	.76
Total Staffing***	4.03	0.82	4.04	0.86	4.03	0.82	.59
% of NH residents on Medicaid	0.61	0.19	0.60	0.20	0.61	0.19	.026
	%	n	%	n	%	n	P
Chain Membership	60%	123,458	59%	5,068	60%	118,373	.15
Ownership: for-profit	64%	130,695	63%	5,383	64%	125,277	.57
Ownership: government	1%	1,986	1%	76	1%	1,892	.33
Ownership: non-profit	35%	72,622	36%	3,130	35%	69,441	.51
Metropolitan	72%	148,524	71%	6,053	72%	142,441	.23
Rural	4%	8,338	4%	356	4%	7,981	.94
Urban	24%	48,323	25%	2,171	23%	46,080	.24
Time
	%	n	%	n	%	n	P
Winter (Dec-Feb)	24%	53,577	26%	2,336	24%	51,204	.012
Spring (March-May)	26%	57,379	27%	2,445	26%	54,903	.019
Summer (June-Aug)	25%	54,766	23%	2,092	25%	52,660	.002
Autumn (Sept-Nov)	25%	54,008	24%	2,154	25%	51,830	.06

Note: AOR, Adjusted Odds Ratio; CI, Confidence Interval.

^*Controlling for resident demographics, health status, NH characteristics, time, seasonality and interactions between registered nurse, licensed professional nurse and certified nurse assistant staffing levels.

^**Significant at alpha = .05

^***measured in full time equivalent hours per resident per day

In the multivariate analyses, maturity of administrative IT capabilities was associated with lower odds of UTI (AOR: 0.906, 95% CI: 0.843, 0.973), controlling for aforementioned covariates (See Table 3). No components of this domain/dimension were individually associated with UTI, nor were any other IT maturity dimension/domain scores. The relationship observed between odds of UTI and maturity of administrative IT capabilities was robust to varying assumptions (data not shown).

Table 3.

Logistic regression of long-stay residents in 816 nursing homes (NH), with NH fixed-effects, state cluster standard errors, survey weights, all domains and controls^*

		AOR	P	95% CI Upper	95% CI Lower
Resident Care	IT Capabilities	1.071	.26	0.950	1.208
	Extent IT Use	0.944	.38	0.830	1.074
	IT Integration	0.975	.67	0.870	1.093
Clinical support Laboratory, Pharmacy, Radiology	IT Capabilities	1.056	.42	0.925	1.207
	Extent of IT Use	0.933	.33	0.812	1.072
	IT Integration	0.941	.24	0.850	1.041
Administrative Activity	IT Capabilities	0.906	.007**	0.843	0.973
	Extent IT Use	1.029	.51	0.945	1.119
	IT Integration	0.953	.16	0.892	1.018

Note: IT, Information technology.

^*Controlling for resident demographics, health status, NH characteristics, time, seasonality and interactions between registered nurse, licensed professional nurse and certified nurse assistant staffing levels.

^**Significant at alpha = .05

Discussion

To our knowledge, this study was first to determine a link between NHs’ IT maturity and health outcomes at the resident-assessment level. In this study, resident characteristics associated with UTI aligned with a previous publication (Castle et al., 2017), though the proportion of regular assessments recording UTI was lower than in previous MDS data (Herzig et al., 2017). Further, the relationship between maturity of IT administrative capabilities and lower odds of UTI was robust. While no individual IT maturity components were associated with the outcome, overall, this type of IT maturity may relieve the administrative burden of tracing UTIs, data documentation and record retrieval affording NH staff time to focus their efforts on improving clinical policy and practice. For example, this section of the survey measures reengineering IT systems such as systems supporting electronic resident care processes and physician order entry (G. L. Alexander et al., 2010). Additionally, network administration, hardware development and software management activities occurring as part of the systems development lifecycle are also measured. These processes are important for IT optimization to enhance efficiency, quality and safety of care delivery (Pollack, 2021). This finding, that administrative IT capabilities are associated with decreased odds of UTI, is additionally robust given healthcare facilities with better documentation systems have potential for measurement bias as they are more likely to record health outcomes, such as UTI.

This work is limited by the nature of the MDS data, which does not identify chronic or drug resistant UTIs. Thus, the same UTI could possibly appear on multiple quarterly assessments, exposing the analysis to bias. Further, this work addresses the relationship between NH structures (i.e., IT maturity) and health outcomes (UTI), controlling for resident characteristics (See Figure). It does not address the relationships between NH processes affecting UTI and IT maturity, which should be the subject of future work.

This work is timely and relevant to policy decisions at facility and public health levels. During the pandemic, all CMS-certified NHs joined the National Health Safety Network (NHSN) to report COVID-19 processes and outcomes. If NHs continue to be active in NHSN, IT maturity is likely to increase, which may improve resident care. Further studies are needed to determine whether lower documentation burden through more sophisticated IT shape clinical practice and policy.