The Effect of the NorGeP–NH on Quality of Life and Drug Prescriptions in Norwegian Nursing Homes: A Randomized Controlled Trial

Background: The effect of the Norwegian General Practice–Nursing Home (NorGeP–NH) criteria has never been tested on clinical outcomes in nursing home (NH) residents. We performed a cluster-randomized trial in Norwegian NHs and tested the effect of NorGeP–NH on QoL (primary outcome), medication prescriptions, and physical and mental health (secondary outcomes) for the enrolled residents; Methods: Fourteen NHs were randomized into intervention NHs (iNHs) and control NHs (cNHs). After baseline data collection, physicians performed NorGeP–NH on the enrolled residents. We assessed the difference between cNHs and iNHs in the change in primary outcome from baseline to 12 weeks and secondary outcomes from baseline to eight and 12 weeks by linear mixed models; Results: One hundred and eight residents (13 lost to follow-up) and 109 residents (nine lost to follow-up) were randomized to iNHs and cNHs, respectively. Difference in change in QoL at 12 weeks between cNHs and iNHs was not statistically significant (mean (95% CI)): −1.51 (−3.30; 0.28), p = 0.101). We found no significant change in drug prescriptions over time. Difference in depression scores between cNHs and iNHs was statistically significant after 12 weeks. Conclusions: Our intervention did not affect QoL or drug prescriptions, but reduced depression scores in the iNHs. NorGeP–NH may be a useful tool, but its effect on clinical outcomes may be scarce in NH residents. Further studies about the effectiveness of NorGeP–NH in other healthcare contexts and settings are recommended.


Introduction
It is well established that polypharmacy, often defined as the use of more than five concomitant drugs [1], is prevalent in nursing homes (NHs) and is associated with frailty, hospitalization, cognitive and physical impairment, falls, and mortality [2,3].
In the past years, several explicit lists, such as Beers Criteria [4,5], START/STOPP [6], EU (7)-PIM [7], the PRISCUS list [8], and NorGeP [9], have been introduced to identify potentially inappropriate medications (PIM) in older adults. Medications may be inappropriate when their potential harm exceeds their benefit [10]. Over 30 PIM lists have been published between 1991 and 2017 aiming to identify the complexity of drug therapy in older people, but these lists have wide variability in what is considered a PIM [11]. This about NorGeP-NH, after receiving a lecture on psychotropic drug use in older adults, and requesting them to perform a structured medication chart review with NorGeP-NH. As for secondary outcomes, we examined whether the same intervention influenced PTD and total medication prescriptions, cognitive function, NPS, physical health status, and functioning in daily living in the same group of residents.

Trial Design
We performed a cluster-randomized trial in 14 NHs, with a total of 42 wards, distributed in eight municipalities in Østfold county, Norway, between November 2018 and June 2019. NHs were treated as clusters, as the intervention was at staff/physician level and not at resident level. Primary and secondary outcomes were at resident level. The NHs were cluster-randomized into two groups, and the NHs were given the name of intervention NHs (iNHs) and control NHs (cNHs). Allocation was not revealed to the NH personnel until after completion of baseline data collection in order to minimize detection bias at baseline. Many of the chosen assessment tools needed to be administered by nurses/authorized social workers who knew the participants well and who had observed the participants over time. Thus, it was not possible to blind data collectors after the intervention was delivered. The report of this trial follows the recommendations of CONSORT (Consolidated Standards of Reporting Trials) guidelines and CONSORT extension to cluster-randomized trials [34]. The described intervention follows the TIDieR criteria [35].
Every participant gave written informed consent to be included in the trial. The capacity to consent was evaluated by a clinical examination performed first by the NH physician, and confirmed by the first author, to detect the participant's ability to understand and weigh the given information, reason, and give an explicit choice. In case of doubt, clinicians could use the Aid to Capacity Evaluation (ACE) form [36]. If participants had reduced capacity to consent, a written informed consent was obtained from the participant's next of kin. The Regional Committee for Medical and Health Research Ethics (2017/2171 REK south-east D) approved the trial. The study was registered on 6 November 2018, on clinicaltrials.gov (accessed on 9 January 2022) (NCT03736577).

Participants
Before inclusion, all 19 municipalities in the district served by the regional Østfold Hospital, with a total of 34 long-term care NHs, received information about the study protocol and were invited to participate. Those responsible for healthcare services in every municipality decided which nursing home(s) could participate. Once the participating NHs were determined, the responsible NH physicians were informed about eligibility criteria to include the NH residents in this study. Eligibility criteria were (a) NH resident and (b) expected to live in the NH for more than 12 weeks. Exclusion criteria were (a) terminal disease, (b) severe somatic or psychiatric disease where the resident was too debilitated or not able to cooperate or where the examination would cause too great of a psychological and physical burden (i.e., severe psychotic state), and (c) the physician had performed a structured drug review for the participant within three months prior to inclusion. NH physicians were thoroughly informed about these criteria and were responsible for assessing eligibility.
Prior to baseline data collection and randomization, the healthcare personnel from both iNHs and cNHs participated in a three-hour lecture on dementia and dementiarelated neuropsychiatric symptoms, delirium, depression, anxiety, and psychosis in older people. In addition, we asked each participating nursing home to dedicate one or two NH personnel per ward to collect data. The data collectors were nurses or authorized social workers, and they participated in a three-hour lecture to learn how to use validated assessment tools. The assessment tools were either interview-based or proxy-based, and they are described later. The first author gave both lectures. Clinical data about the residents in both iNHs and cNHs were collected eight and 12 weeks after baseline data collection.

Intervention
The intervention was an educational intervention on nursing home physicians, followed by a drug chart review of the participant's medications, and included the following steps: (1) Physicians in the iNHs attended a three-hour lecture including the following subjects: principles of pharmacology in older people; -the use of PTDs in older people; -how to conduct a drug chart review with the Norwegian General Practice-NH (NorGeP-NH) criteria [27].
This lecture was held by a psychiatrist (first author) after baseline data were collected. The lecture was held in the nursing home where the physicians worked. It was held face-to-face and included an electronic presentation as supportive material. The physicians who attended the lecture were given a copy of the electronic presentation after the lecture, and they received a laminated NorGeP-NH list to use in the following step.
(2) Within a two-week period after the lecture, physicians in the iNHs performed a drug chart review according to NorGeP-NH. Physicians were allowed to consult a psychiatrist (first author) in case they needed to discuss choices made during a review, but the final decision about medication changes was the physician's responsibility.

Control Group
The physicians and healthcare personnel in the cNHs were asked to follow-up residents as usual. If medication changes were necessary, physicians could do so, but without using a structured drug review chart during the follow-up period. After the last assessment at 12 weeks, as a courtesy to the physicians in the cNHs they were given the same lecture as described in (1).

Collected Data and Outcomes
The primary outcome was the difference between cNHs and iNHs in change in quality of life (QoL) assessed with the Quality of Life in Late-Stage Dementia (QUALID) scale [30,37], from baseline to 12-week follow-up. The secondary outcomes were the difference between cNHs and iNHs in change from baseline to 8-12 weeks in the number of drugs prescribed daily, the number of prescribed pro re nata (PRN) drugs, the prescription of psychotropic drugs categories (antidepressants, antipsychotics, anxiolytics, sedatives/hypnotics, and antidementia drugs), and in clinical scores measuring the level of depression, cognitive function, neuropsychiatric symptoms, physical health status, and functioning in daily living. Table 1 reports the instruments used to collect the data. We also collected demographic data and nursing home characteristics for each participating resident ( Table 2 in Results section).  Proxy-based 0-144 Single-item score is calculated by multiplying severity (score 1-3) by frequency (score 1-4). Total score is the sum of all the single-item scores [40][41][42]. We calculated the NPI-NH subsyndrome scores for agitation, psychosis and affective symptoms b .
Cornell Scale for Depression in Dementia (CSDD) Total score is calculated by summing up 19 single-item scores. Each single item can be scored 0, 1 or 2 (symptom not present, moderate or periodically present, severe). A higher score indicates more severe symptoms [43].

Anatomic Therapeutic
Chemical (ATC) classification system Medication chart in resident's journal N/A We calculated the total amount of daily prescribed drugs, and the total amount of prescribed pro re nata (PRN) drugs. We collected data on prescribed psychotropic drugs, and we grouped them as antipsychotics (N05A except lithium), antidepressants (N06A), anxiolytics (N05B), hypnotic/sedatives (N05C), and anti-dementia medication (N06D

Sample Size
In a previous Norwegian study, people admitted to an NH had a QUALID score of 20.0 (SD 7.2) [51]. When the study was designed, to the best of our knowledge, there were no previous randomized controlled studies using QUALID score as a primary outcome. Thus, to be sure that any possible change caused by our intervention was clinically relevant, we chose to define a change from baseline to 12-week follow-up in QUALID score of 33% as of clinical importance prior to power calculations. With an 80% power and 0.05 significance level, and assuming SD 7.2 in both groups at baseline and follow-up, 39 residents needed to be included in each allocation group to detect a 33% difference between iNHs and cNHs in change from baseline to 12-week follow-up in QUALID score. In Norwegian NHs, about one out of four residents die every year [52]. Thus, we estimated a 6-7% drop-out rate due to death within a 12-week period. Rounding up the drop-out rate to 10%, 43 residents had to be included in each allocation group. Assuming 10 participating NHs and a cluster effect on NH level of about 5%, the final number of residents was estimated to be 60 in each group. Because of uncertainty about how many NHs would decide to participate in the study, we aimed to include about 100 residents in each allocation group.

Randomization
An independent statistician allocated the participating NHs into two arms by performing a stratified randomization using a computer-generated algorithm. To avoid contamination bias, every NH was treated as a cluster. Each NH was under the care of one physician or group of physicians who worked together and only in that NH. All the participating NHs were stratified into four groups. Stratification was performed according to the number of participants the personnel in each NH were able to include and follow up. The allocation results were kept in a digital safe, hidden from NH physicians responsible for enrolling participants. NH physicians were asked to assess each resident in the participating NH for eligibility, and they were responsible for enrolling participants. One of the authors (EC) verified the eligibility criteria by discussing them with the NH physician/NH personnel and verified the participants' capacity to consent. If an NH had limited resources to follow up participants, NH leaders and physicians were asked, before inclusion and allocation, to determine how many residents they could possibly enroll and follow up. In this case, the predetermined number of residents was selected by drawing lots. This process was performed by EC in the presence of at least one healthcare personnel from the selected NH.
Once NHs were allocated and residents were enrolled, a random-number generator was used to determine which allocation group was given the intervention. The result of this process was also kept hidden from NH physicians and healthcare personnel until after baseline data were collected. Once baseline data were collected, the first author informed the physicians who were working in the intervention NHs and carried out the intervention together with them.

Statistical Methods
The statistical analyses were performed by using SPSS© v27 and SAS© v9.4. Baseline characteristics are presented as means and standard deviations (SDs) for continuous variables and frequencies and percentages for categorical variables. We present the total amount of prescribed drugs as means (SDs) both for daily prescriptions and for pro re nata (PRN) drugs. PTD prescriptions are presented as frequencies and percentages at each assessment point. For the primary and secondary analyses, we included participants who had data available at baseline. To assess the difference between iNHs and cNHs in the change in primary and secondary continuous outcomes, we estimated linear mixed models with fixed effects for time, allocation group, and interaction between these two. To assess the difference in change for categorical outcomes, we estimated generalized linear mixed models with the same fixed effects. All models contained random effects for NHs to adjust the estimates for cluster effect at the NH level, which was non-negligible according to the intra-class correlation coefficient. For continuous outcomes, the results were presented as mean change within allocation groups and mean difference in change between the groups with corresponding 95% confidence interval (CI) and p-value. For categorical outcomes, the results were presented as odds for change within the allocation group as well as odds for differences in change between groups with 95% CI and p-values. We set the level of significance at 5%. Figure 1 shows the flow diagram of the trial. Two hundred and seventeen residents were included at baseline between November 2018 and March 2019. Six hundred and three residents from 15 NHs in the nine municipalities that agreed to participate in the trial were recruited and assessed for eligibility. Among these, 437 met inclusion criteria. One NH in one municipality, which had originally agreed to participate with 14 residents, withdrew from the trial during eligibility assessment due to a lack of local NH resources. Fifty residents declined to participate, 161 residents were excluded by drawing lots because some NHs could not include more than a predetermined number of residents (see Section 2.7), six residents died right before baseline assessment, and one resident moved from the NH right before baseline assessment. Two residents were excluded for violation of protocol, as the NH never returned the assessment documentation. Sixteen NH physicians were involved in the trial, seven working in the cNHs and nine working in the iNHs. Table 2 reports demographics, NH characteristics, and clinical scores at baseline. Residents were on average (SD) 84.6 (9.4) and 83.3 (8.0) years old in the cNHs and iNHs, respectively. Most residents in the control group lived in regular units (56.9%), while most residents in the intervention group lived in special care units (59.3%). The two groups had a comparable number of residents per unit (15.07 in cNHs vs. 13.15 in iNHs), number of staff members per unit during the day shift (4.73 in cNHs vs. 4.61 in iNHs), and physicians worked on average 0.88 more hours in cNHs compared to iNHs (6.43 h in cNHs vs. 5.55 h in iNHs). According to CDR, most participants had either mild cognitive impairment (7.8% in cNHs and 7.7% in iNHs) or dementia (89.3% in cNHs and 92.3% in iNHs). The average (SD) number of prescribed daily drugs was 6.92 (3.49) for participants living in the cNHs and 7.55 (3.04) for participants living in the iNHs. The average number (SD) of prescribed pro re nata (PRN) drugs was 4.04 (2.74) for the cNHs and 4.72 (2.89) for the iNHs. For some of the assessment scores, such as MoCA, CSDD, QUALID or TUG, there was a considerable amount of missing data. This aspect is discussed later.

Results
Results from the primary analysis, assessing the difference in change in QoL, are presented in Table 3. We found no statistically significant difference between cNHs and iNHs in change in QoL from baseline to 12-week follow-up. However, while the QUALID score remained stable in the iNHs, we found a statistically significant increase in QUALID score (higher QUALID score indicates lower QoL) in the cNHs (p = 0.013).   Results from the analyses of secondary outcomes (see paragraph "Outcomes" and Table 1 for details) are presented in Table 4 for clinical measures and Table 5 for prescribed drugs. Compared to the control group, residents in the iNHs had a significantly larger reduction in CSDD score from BL to week 12 (mean difference in change (95% CI) −2.59 (−3.95; −1.23), p < 0.001). We found no statistically significant difference between the two groups in change in the prescription of PTD categories (antidepressants, antipsychotics, anxiolytics, sedatives/hypnotics, and antidementia drugs treated as groups).      Further, compared to the control group, participants in the iNHs had a statistically significant reduction in GAI score from BL to week 8 (−1.69 (−3.37; −0.01), p = 0.049), and a statistically significant reduction in the total amount of prescribed daily medications from BL to week 8 (−0.41 (−0.75; −0.06), p = 0.023). Residents in the iNHs, compared to residents in cNHs, had a significantly larger reduction in the odds of having a lower CDR score from baseline to week 8 (p = 0.007), but no significant difference in reduction from baseline to week 12.

Brief Synopsis of Key Findings
This NH trial examined how an educational program on psychopharmacology and the use of NorGeP-NH in a real-world setting influenced QoL, other clinical outcomes, and medication prescriptions among NH residents. Our trial did not show any significant difference in change in QoL scores between iNHs and cNHs from BL to 12-week follow-up. Even though we found a statistically significant reduction in QoL among cNHs residents from BL to 12-week follow-up, this reduction was not relevant from a clinical perspective, according to what we assumed to be a clinically important reduction (33%). Our trial showed that the intervention did not reduce the total amount of daily prescribed drugs in the iNHs compared to cNHs after 12 weeks. However, there was a significant, yet temporary, reduction of the total amount of daily prescribed drugs from BL to 8 weeks, in the iNHs. Our results also showed that the depression score was significantly lower in the intervention group compared to the control group, at 12-week follow-up. Our trial did not show any significant difference between cNHs and iNHs in change for PTD category prescriptions.

Strengths and Limitations
A strength of this study is the fact that participants were assessed by healthcare personnel who worked in the NHs where the participants lived and who had good knowledge of the participants' clinical history. We also chose to focus on a "real-world" intervention performed by the same physicians treating the participants in the usual care setting, and not by conducting the intervention by external personnel, which has been previously discussed and has shown a possible lower adherence to suggested medication changes [22,25,26].
Another strength of our study is that NH personnel performed clinical evaluations with validated tools, commonly used both in Norway and internationally in NHs. This makes it easier to compare our results with other studies. In addition, only one investigator (first author) directed the intervention and follow-up assessments, and by having direct contact with every participating NH, the possibility of missing data in the dataset was minimized.
This study has several limitations. Participants were cluster-randomized to minimize within-NH contamination bias. Despite cluster-randomization, the intervention and control groups may have had differences that impacted the results. For example, at baseline, a higher number of participants in the iNHs lived in special care units. This might reduce the potential beneficial effects of the performed intervention due to a higher level of morbidity in residents admitted to special care units, and, as a consequence of that, a lower QoL over time. Further, the short duration of the trial might not have been long enough to assess the long-term effects of the intervention. However, due to a high mortality rate in NHs [52], a shorter trial duration may reduce the number of people dropping out of the study.
Some assessment scores at BL, such as MoCA, QUALID, CSDD, GAI, or TUG, had several missing data. This may cause uncertainty when comparing the groups. We do not have an explanation for the reason why there were many missing data for QUALID or CSDD, as they are proxy-based assessment tools. For MoCA, GAI, and TUG, a possible explanation is that they require direct cooperation of the residents, which may have been difficult to achieve due to severe cognitive or physical impairment.
The data collectors were nursing home personnel, and they were blinded only during baseline data collection. This is a possible source of detection bias, as the assessors may have been influenced by the knowledge of randomization. However, most of the proxybased assessment tools needed a deep knowledge of the participants and an observation time that lasted several days or weeks before an assessment was performed. Therefore, it would also be problematic to have the participants assessed by external, fully blinded personnel. Further, we did not analyze inter-rater reliability, and this may have led to bias in the data collection process, due to possible differences between data collectors. We did not collect precise data about how many nurses/authorized social workers participated in the data collection process, as this was decided internally in each nursing home ward. However, 42 wards participated in the study, and each ward had at least one or two data collectors. A large number of data collectors may also reduce a possible skewing of the distribution in the collected data.
Finally, we did not assess potential economic consequences of the intervention, which might be important to support such educational interventions in the future.

Considerations and Comparison with Relevant Studies
Our trial did not show a significant difference between the two groups in change in QoL. It is possible that our educational intervention, which focused on medication review and was only performed once, was not enough to improve QoL in the short term. In fact, a multicomponent, long-lasting intervention conducted in Norway showed no change in QoL during the first four months of intervention, but it showed an improvement nine months later [53]. However, our results are still in line with a Cochrane review conducted in 2016. That review analyzed the effect of different interventions to optimize drug prescriptions in NH residents and found no strong evidence showing intervention efficacy on residentrelated outcomes, such as hospital admissions, mortality, adverse drug events, or QoL [16]. It is still important to note that even though we found no significant difference between the two groups in change in QoL, QoL did not worsen in the intervention group, while there was a significant reduction in QoL for participants in the cNHs. Indeed, the intervention might have prevented a possible worsening in QoL for the iNHs. However, the worsening in QUALID score we found in the control group may not be clinically relevant, as the mean (SD) score worsened from 21.31 (6.72) to 22.74 (7.64).
This trial showed a significant reduction in the total amount of prescribed drugs eight weeks after intervention, but not after 12 weeks. A possible explanation is that some drugs might have been reintroduced after a temporary discontinuation. However, we have not analyzed changes for all types of medication, as we only examined drug categories in the N05-/N06-ATC groups in detail. Our results seem to find partial support in previous studies. A recent systematic review showed that medication reviews can improve the appropriateness of drug prescription [17], which in some cases requires drug discontinuation. However, our trial may show that the effect of a medication review on the total medication amount is only temporary. The reduction in the total amount of drugs prescribed daily for older residents may have a beneficial effect on several clinical factors, such as reduction of frailty, improvement in cognitive function, or lower risk of falls [2]. However, among the clinical outcomes examined in our trial, only depression improved after 12 weeks. On the one hand, depression is an important risk factor of polypharmacy and excessive polypharmacy [54]. On the other hand, several somatic prescription medications, such as antihypertensives, proton pump inhibitors, or analgesics, are known to possibly cause depression as an adverse outcome [55].
Residents living in iNHs showed an increase in CDR score during follow-up. A possible explanation might be that more participants in the iNHs lived in special care units, often offered to people with moderate to severe dementia symptoms and poor prognosis. This difference might also explain why at BL residents living in iNHs presented more severe cognitive symptoms measured by CDR and MoCA as well as more severe NPS measured by NPI-NH.
Even though a recent systematic review and meta-analysis showed that focused psychotropic medication review is effective in reducing PTD prescriptions in NH residents [22], our intervention used a more general drug chart review tool which is not PTD-specific, and this might explain why we did not find a significant difference in change in PTD prescriptions between the two groups. NorGeP-NH is described in a recent review as limitative, as it does not include possible inappropriate medications for specific comorbidities [56]. However, a newly performed multidisciplinary, long-term NH cluster-randomized intervention, using other drug review lists, also failed to show an effect in reducing PTD prescriptions [23].

Conclusions
Our intervention on the use of NorGeP-NH and on the correct use of PTD in older NH residents did not have an effect on QoL or PTD prescriptions in the short term. However, our intervention still showed a positive, yet temporary, effect on the total drug load residents received and on the level of depression. NorGeP-NH may still have value in clinical practice, even if the evidence of its beneficial clinical effects may be scarce. Future research on the use of NorGeP-NH and other medication review tools should be performed in NHs and in other clinical settings to assess their real effectiveness on medication prescriptions and on the overall health status of older adults.