Association of Infection-Related Hospitalization With Cognitive Impairment Among Nursing Home Residents

This cohort study uses data from the Minimum Data Set and Medicare hospitalization records to assess the association between infection-related hospitalization and cognitive function among nursing home residents 65 years or older.


Section 3. Statistical model
To visually (and flexibly) assess the pattern of cognitive outcomes relative to the quarter of InfHosp, we start by estimating the following event study model: Where Yijt describes the outcome of interest for NH resident i, living in a nursing home j at time t, denoting calendar years and quarters. The subscript q refers to event quarters, which are quarter intervals relative to the quarter prior to InfHosp (i.e., period q = -2 refers to 2 quarters before the InfHosp), and % are coefficients on indicators for quarters relative to the InfHosp. Thus, this model effectively compares changes in cognitive outcomes in each quarter increment before vs after InfHosp for each NH resident -that is, it estimates the differences in outcomes for leads and lags of InfHosp relative to a reference pre-hospitalization quarter (t -1) -these relative differences are described by the coefficients % for each q. We presented these coefficients in all of our main and supplement figures, along with 95% confidence intervals, adjusted for serial correlation in the outcome at the NH level. Controls for resident's age, coexisting medical conditions (i.e., diabetes, cancer, ischemic heart disease) as well as a binary variable indicating hip fracture or shortness of breath were also included ( !# * ). Calendar quarterly and year indicators, # , were used to adjust for © 2021 Gracner et al. JAMA Network Open.
nationwide secular trends and seasonality in the cognitive outcomes (see Figure 2 and eTable 2, Guided by the patterns obtained using the model (1), and to summarize the magnitude of estimated effects and their statistical significance, a linear pre-trend in event time was allowed and superimposed, as described in our model (2) below: (2) !"# = $ + + ∑ % + !# * + ! + # + !"# ) %($ Model (2) allowed for a linear pre-trend in event time q (i.e., between quarter MDS assessments).
We decided for the linear trend based on the Figure 2 based on model (1), which suggested that a linear trend captures secular trends well. The key coefficient of interest here was % , which measured the change in cognitive function following the InfHosp relative to any pre-existing linear trend ( ). In this model (2), the identifying assumption was that, conditional on controls and having an InfHosp, the timing of the InfHosp was uncorrelated with deviations of the outcome from a linear trend in event time. Finally, because we observed that transfer effects on outcomes were mostly immediate and persisted over time, we also estimated an average effect of all combinations of comparisons between residents across all post-transfer quarters by replacing ∑

Section 4. Sensitivity Analyses.
A major concern in our analyses is non-random attrition, arising primarily due to mortality. It poses a potential threat to our identifying assumption if it is correlated with our outcome of interest, in either levels or changes. This is likely, if hospitalizations for NH residents with worse prehospitalization cognitive outcomes are also more likely to result in death. eFigure 3 shows that a large share of NH residents indeed died within the first quarter since InfHosp. eTable 1 compares residents who died in the quarter or within the first three quarters of InfHosp and shows that though on average similar in characteristics like race and co-morbidities, residents who died in the quarter of InfHosp had worse cognitive function prior to InfHosp than did the long-term InfHosp survivors (e.g., 13.7% vs 9% with severe cognitive impairment). They were also older (84.4 vs 82.7), less likely to be female (64.1% vs 70.9%), and majority were transferred due to sepsis: 67% vs 41%.
When restricting our study sample to those who died within one year of transfer, our estimates show significant changes across all cognitive outcomes (see eTable  within-individual changes and therefore should not be contaminated by any differential attrition correlated with the level of the outcome. We further investigated the potential concern regarding the compositional changes in the set of individuals used to identify our estimates by re-estimating our specifications on a balanced panel of residents observed in at least 6 quarters prior and after the InfHosp. Estimation on this balanced panel allows us to examine the time pattern of outcomes without concerns about potential effects of compositional sample changes (eTable 4).
Finally, to address the concern of spurious correlation between InfHosp and cognitive decline, we also reproduced the analysis under conditions in which it was unlikely to demonstrate our findings.
Specifically, we assigned random placebo hospitalization dates (e.g., we changed hospitalization dates to 4 quarters back from their actual date). We found no significant relationship between them and outcomes of interest. See eTable 5 for results.

Section 6. Infection-related hospitalization and delirium symptoms.
We followed the Confusion Assessment Method (CAM) criteria (validated with high sensitivity, specificity and reliability in previous work [6][7][8][9] when identifying residents with delirium. Specifically, following Kosar et al. 9 , we defined delirium as present if patient was reported to have demonstrated an acute change or fluctuation in delirium symptoms, such as inattention, and either disorganized thinking or an altered level of consciousness (i.e., MDS variables c1300a/b/c/d equal to 2, indicating one or more delirium signs and symptoms that fluctuate present), or if there was evidence of an acute change in mental status from the resident's baseline (c1600 equals 1). eFigure 5A below shows that increased prevalence of delirium symptoms post-InfHosp is short-lived: we observe a significant increase of about 1 percentage point in fluctuating delirium symptoms during the first quarter post-InfHosp that return to their pre-hospitalization trajectory after that. In contrast, following InfHosp delirium symptoms that are both, continuously present or fluctuating (defined with the MDS variables above, and if variables c1300a/b/c/d equal to 1; indicating that one or more delirium signs and symptoms are continuously present), increase permanently. Specifically, we found that even one year after InfHosp, the probability of experiencing delirium symptoms that are continuously present is higher by nearly 1.5 percentage points than it would have been in the absence of InfHosp (eFigure 5B). Quarters relative to hospitalization Notes: Changes in cognitive outcomes over time by subgroups. Estimated models are identical to those in Figure 2, except that here each model is estimated for each subgroup. Panel A, older or younger than 85 years; Panel B residents hospitalized due to sepsis vs other infection-related condition. See Figure 2 caption for further details.   Notes: a) Reference year is 2011. b) Reference quarter is Jan-March. c) Reference age group is 65-69. d) Reference group are those without ADRD or with 0 years with the disease. Chronic conditions include diabetes, hypertension, obesity and stroke or TIA. All regressions include NH resident indicators. **, and *** refer to a p-value smaller than 0.05 and 0.001, respectively. Robust standard errors are clustered at the nursing home level. 95% confidence intervals are reported in brackets. OLS refers to ordinary least square and LPM to linear probability model estimation approach.     Carbuncle and furuncle of hand 680. 5 Carbuncle and furuncle of buttock 680. 6 Carbuncle and furuncle of leg 680.7 Carbuncle and furuncle of foot 680. 8 Carbuncle and furuncle of head and scalp 680. 9 Carbuncle and furuncle unspecified site 681.00 Cellulitis and abscess of finger, unspecified 681. 10 Cellulitis and abscess of toe, unspecified 681. 9 Cellulitis and abscess of unspecified digit 682.0 Cellulitis and abscess of face 682. 1 Cellulitis and abscess of neck 682. 2 Cellulitis and abscess of trunk 682. 3 Cellulitis and abscess of upper arm and forearm 682. 4 Cellulitis and abscess of hand, except fingers and thumb 682. 5 Cellulitis and abscess of buttock 682. 6 Cellulitis and abscess of leg, except foot 682.7 Cellulitis and abscess of foot, except toes 682. 8 Cellulitis and abscess of other specified sites 682. 9 Cellulitis Pneumonia due to other Streptococcus J15. 5 Pneumonia due to Escherichia Coli (E. Coli) J15. 6 Pneumonia due to other Gram-negative bacteria J15. 7 Pneumonia due to mycoplasma pneumonia J15. 8 Pneumonia due to other specified bacteria J15. 9 Bacterial pneumonia unspecified J16.0 Pneumonia due to chlamydia J16. 8 Pneumonia due to other specified organism J18. 1 Lobar pneumonia, unspecified organism J20.0 acute bronchitis due to mycoplasma pneumoniae J20.2 acute bronchitis due to streptococcus J340 abscess, furuncle and carbuncle of nose J36 Peritonsillar abscess J38.7 Cellulitis and perichondritis of larynx J39. 1 Other abscess of pharynx J85.0 Gangrene and necrosis of lung J85. Other and unspecified infection due to central venous catheter T81.4XXA Infection following a procedure, initial encounter T82.6XXA Infection and inflammatory reaction due to cardiac valve prosthesis, initial encounter T82.7XXA Infection and inflammatory reaction due to other cardiac and vascular devices, implants and grafts, initial encounter T83.51XA Infection and inflammation due to indwelling urinary catheter T83.590A Infection and inflammatory reaction due to implanted urinary neurostimulation device, initial encounter T83.591A Infection and inflammatory reaction due to implanted urinary sphincter, initial encounter T83.592A Infection and inflammatory reaction due to indwelling ureteral stent, initial encounter T83.593A Infection and inflammatory reaction due to other urinary stents, initial encounter T83.598A Infection and inflammatory reaction due to other prosthetic device, implant and graft in urinary system, initial encounter T83.61XA Infection and inflammatory reaction due to implanted penile prosthesis, initial encounter T83.62XA Infection and inflammatory reaction due to implanted testicular prosthesis, initial encounter T83.69XA Infection and inflammatory reaction due to other prosthetic device, implant and graft in genital tract, initial encounter T84.50XA Infection and inflammatory reaction due to unspecified internal joint prosthesis, initial encounter T84.60XA Infection and inflammatory reaction due to internal fixation device of unspecified site, initial encounter T84.7XXA Infection and inflammatory reaction due to other internal orthopedic prosthetic devices, implants and grafts, initial encounter T85.71XA Infection and inflammatory reaction due to peritoneal dialysis catheter, initial encounter T85.730A Infection and inflammatory reaction due to ventricular intracranial (communicating) shunt, initial encounter T85.731A Infection and inflammatory reaction due to implanted electronic neurostimulator of brain, electrode (lead), initial encounter T85.732A Infection and inflammatory reaction due to implanted electronic neurostimulator of peripheral nerve, electrode (lead), initial encounter T85.733A Infection and inflammatory reaction due to implanted electronic neurostimulator of spinal cord, electrode (lead), initial encounter T85.734A Infection and inflammatory reaction due to implanted electronic neurostimulator, generator, initial encounter T85.735A Infection and inflammatory reaction due to cranial or spinal infusion catheter, initial encounter T85.738A Infection and inflammatory reaction due to other nervous system device, implant or graft, initial encounter T85.79XA Infection and inflammatory reaction due to other internal prosthetic devices, implants and grafts, initial encounter A04.72 enterocolitis due to clostridium difficile, not specified as recurrent Z16. 10  resistance to multiple antimycobacterial drugs A08. 8 Other specified intestinal infections J18.8 Pneumonia organism unspecified J18. 9 Pneumonia, unspecified organism J20. 9 Acute bronchitis J22 unspecified acute lower respiratory infection J44.0 Chronic obstructive pulmonary disease with acute lower respiratory infection L04. 9 Acute lymphadenitis T80.22XA Acute infection following transfusion, infusion, or injection of blood and blood products, initial encounter T80.29XA Infection following other infusion, transfusion and therapeutic injection, initial encounter T83.511A infection and inflammatory reaction due to indwelling urethral catheter, initial encounter T83.518A infection and inflammatory reaction due to other urinary catheter, initial encounter T88.0XXA Infection following immunization, initial encounter