Additional research into the psychometric properties of the PDRQ-9 was recommended to investigate its reliability and validity in larger primary care or general hospital samples (1) and compare it to a measure assessing the patient-doctor relationship from the doctor's standpoint, such as the 10-item Difficult Doctor-Patient Relationship Questionnaire (DDPRQ-10) (19). Another recommendation was to explore whether the PDRQ-9 test scores correlate with clinically important outcomes such as higher clinical follow-up rates, patient compliance with treatment recommendations, illness outcomes, or functional status (1). Since its publication, this questionnaire has been adopted widely in the literature on the PDR. It was cited over 1,500 times in Google Scholar and has been translated, psychometrically tested and used globally for research-related evaluations. This abundance of publications on the PDRQ-9 enables us to explore some research questions regarding the abovementioned psychometric properties, and the cross-cultural validity of the PDRQ-9. Cross-cultural validity refers to whether measures that were originally generated in a single culture are applicable, meaningful, and thus equivalent in another culture (20). It has been mainly applied in psychological studies which need to adapt self-reported health status measures for use in languages other than the source language (21).

We aim to provide an overview of the developments with the PDQR-9 over the last 20 years, to summarise the findings, and to illustrate for what purpose the PDRQ-9 can be used. Furthermore, we propose a method for normalising scores and signposting areas of research that might be relevant to explore in the future.

Studies presenting data on the PDRQ-9 that used 9-item versions of the PDRQ-9 and the scoring method of those items were eligible for inclusion. Studies were excluded if they were not peer-reviewed, did not provide outcome data for the PDRQ-9, did not use the recommended scoring method or 9 items, developed another questionnaire based on the PDRQ-9, were a protocol publication, or were retracted.

We conducted a search for articles in Pubmed, Medline, Web of Science, and Google Scholar, with the last date at the end of October 2025.

We used the terms “patient-doctor relationship questionnaire” OR “PDRQ-9” OR “PDRQ9” OR “PDQR” (as some studies had reversed the letter order of the abbreviation) published between the publication date of the original study (1) in March 2004 up to November 2025. No language restrictions were used. We checked studies for references to find other studies fulfilling the inclusion criteria. In addition, we consulted a database of PDRQ-9-related studies that was put together over the years by one of the authors (CFC). The search string is mentioned at the end of the document.

Search and screening were conducted independently and in duplicate by the two authors (CFC or EdB) and an assistant, and in case of doubts, they were discussed until consensus was obtained. We outlined the selection of studies in a Flowchart.

All performed data extraction, and in case of doubt, they discussed it until a consensus was reached. We queried 16 authors for additional data or clarification. Six authors (17, 22–26) responded and provided the requested information.

We extracted data from the studies exploring countries and medical settings where the studies were conducted, the type of healthcare professionals who were evaluated with the PDRQ-9, patient populations in which the PDRQ-9 was administered, the method of language translation, if the translated version was provided, any validation or reliability outcomes, scoring method, and any outstanding issues relevant for further research mentioned in the publications. We also summarised findings based on the PDRQ-9 in studies where an already psychometrically tested version was used, like Mean, SD, total scores, and cut-off levels.

The authors (CFC, EdB) assessed the risk of bias using an adapted tool for cross-sectional studies (27), focusing primarily on selection (description target group), sampling method (representative or convenience sample), valid measurement of PDRQ-9 outcome, and selective reporting. Both authors discussed the criteria and assessed the studies in duplicate, examining cases of doubt together.

M and SD, cut scores, norm scores.

We present psychometric and factor structure findings and total scores, calculating weighted Means and SD__pooled over studies. We present an approach to convert raw sum scores to standardised, normalised scores. The most recent guideline on reporting transcultural validity states that it is not only important to consider semantically correct translation and conceptual equivalence. Testing of reliability and construct validity, psychometric analysis and field studies should also be conducted (28). We have structured the findings accordingly. Therefore, Table 2 reports on translation method, such as forward-backward translation (40) and psychometric findings, such as any reliability or validity findings; Table 1 reports on factor analysis, and Table 3 shows field studies with Means and SDs. Weighted mean item scores and SD__pooled are then reported in Table 4. Cut-off values for the quality of the patient-doctor relationship based on normalised T-scores are reported in Table 5. Mean scores are presented in Supplementary Table A1 of the Supplemental Materials. A crosswalk table and formulas for converting PDRQ-9 raw scores into normalised T-scores and Percentile scores is presented in Supplementary Table A2 in the Supplemental Materials.

Some authors sum the item responses to obtain a total score on the PDQR-9, while others calculate the mean score of the items. To present scores in the tables on the same metric, we multiplied the latter mean scores and SDs by 9 to obtain a total score. We calculated weighted Mean Scores and SD__pooled as an estimate of the SD in the total group, taking the N of the study into account, and produced a forest plot per study, per language group, and for the total group. As the scores on the PDRQ-9 appeared skewed, with most patients rating their doctors positively, we presented an approach to convert raw sum scores to standardised, normalised scores based on a large German representative sample dataset of the general population (11).

We report the risk of bias in Supplementary Table A3 in the Supplemental Materials.

We asked the authors of the three potentially representative samples for an anonymised subsample of their dataset to explore the possibility of converting raw sum scores to normalised scores and received two samples: the German (11) and the Brazilian (Portuguese) (71) one. The German sample was the only representative general population sample and was used to establish a conversion table and formula, as described below.

PDRQ-9 total scores were first normalised with a percentile rank order-based approach [Rankit normalisation (18, 78, 79) with the RankNorm function in the R RNOmni package version 1.0.1 (80)]. This function first transforms raw scores to percentile ranks, based on the formula PR = (CF -.5*F)/N (CF = the cumulative frequency of the raw score; F = - the frequency of the Raw score; N = the size of the normative sample). Next, the percentile rank order scores were converted to standard scores (Z-scores) with the probit function, resulting in Z-scores with M = 1 and SD = 1). These Z-scores were converted to T-scores with T = 10*Z + 50 to obtain a more convenient metric with a M = 50 and SD = 10 (81). According to the empirical rule for the normal distribution, 68% of the population will score in the range of T = 40–60, 95% in the range of T = 30–70, and 99.7% in the range of 20–80. Once the T scores have been established, the curvilinear relation between raw scores and normalized T-scores can be estimated with non-linear modelling (the Non-linear Least Squares nls function in the R Stats (version 3.6.2) package and the glsnls package version 1.2.0 (https://cran.r-project.org/web/packages/gslnls/gslnls.pdf). The full procedure is described elsewhere (82). A crosswalk table and formula were established to convert raw PDRQ-9 scores to normalised T scores. We followed the Prisma statement guidelines for systematic reviews and report the PRISMA checklist in the Supplementary Materials (83).

Regarding settings and patient populations, the psychometric studies were conducted with 4,555 primary care patients (1, 6, 8, 11, 13, 16, 17, 31, 34, 35, 37); n = 2,275 of these were from a representative general population sample (11). Eight studies were conducted in 2,338 general hospital outpatients; one in patients with cancer (9), one in patients with cancer or chronic pain receiving medical cannabis (33); one in patients with type 2 Diabetes Mellitus (10); one in patients of internet hospitals (15), one in surgical patients (30); one in orthopaedic patients (22); one in patients with Continuous Ambulatory Peritoneal Dialysis (39) and one in patients with asthma (24). Three studies were conducted with 623 outpatients of specialised mental health institutions (12), one of them specifically for patients with schizophrenia (7) and one for patients with depression (38). Two studies reported on 258 general hospital inpatients: one specifically on neurology, gastroenterology, obstetrics and gynaecology inpatients (14), and one on inpatients in a specialised epilepsy clinic (1). Furthermore, psychometric properties of the PDRQ-9 were evaluated in 82 parents of children presented at a paediatric clinic (29) and in 190 online patients in the general population with fibromyalgia (32).

Construct validity was evaluated by comparing the PDRQ-9 score in different treatment settings. Primary care patients scored significantly better than epilepsy inpatients of a specialised epilepsy clinic (mean difference = .46, p = .001) (1). Primary care patients who were treated by faculty physicians scored higher than those who were treated by resident family doctors (p = .01) (6). There was a correlation between the PDRQ-9 score and duration of treatment (r_pmcc = 0.114, p = 0.044) and number of previous hospitalizations (r_pmcc = 0.179, p = 0.002), and a negative correlation with the Morisky Adherence Scale (r_pmcc = 0.174, p = 0.002) in outpatients with stable schizophrenia (7) and the PDRQ-9 score was found to be significantly predicting adherence to medical cannabis in patients with chronic pain visiting a clinic for medical cannabis prescription (p < 0.01) (33).

Several studies compared the PDRQ-9 score with another questionnaire for the PDR, such as the Difficult Doctor-Patient Relationship Questionnaire (DDPRQ-10), a scale that assesses the doctor's view of the PDR, with which the PDRQ-9 score was correlated negatively (r = −.22, p = .003) (6); the Trust in physician scale (r = 0.413, p < 0.001), with which it correlates positively (7); and satisfaction with pain treatment (r = .51) (11). These correlations suggest moderate, acceptable construct validity.

All studies found the questionnaire reliable, with Cronbach's alpha ranging from 0.62 to 0.97 (mean 0.91). Test-retest reliability ranged from 0.61 over two months to 0.74 after seven days.

Table 1 provides factor structure findings of 14 studies, including 11 of the validation studies (1, 7–17, 29, 135).

All studies confirmed the one-factor model of the PDRQ-9 except one study, which found a two-factor solution for the Chinese version. The first factor reflects empathy and trust experienced by the patient during treatment, and the second factor represents the patients' appraisal of the process and results of their treatment (14).

Table 3 provides the total scores of the PDRQ-9 and other findings from 62 publications describing field studies conducted with psychometrically tested versions of the PDRQ-9. They involved 25,865 patients from 24 countries: Austria, Bangladesh, Brazil, Canada, China, Germany, Ethiopia, Iran, Israel, Italy, Malaysia, the Maldives, Mexico, Nigeria, Oman, Romania, Saudi-Arabia, Spain, Taiwan, Thailand, the Netherlands, Turkey, the United Kingdom and the USA. All studies conducted are shown; in case studies do not show total scores, weighted Mean and SDs are reported.

Most studies were based on convenience samples in clinical settings and one Chinese study reporting on a convenience sample derived from an online survey conducted via internet and social media (75). In addition, some studies were performed in representative samples. Four publications covered one representative general population sample in Germany (11, 61–63), one study used a systematic random sampling method resulting in 372 surgical patients in Ethiopia (30) and one Brazilian study was nested within a national cross-sectional study conducted to assess the quality of care provided by the Brazilian Primary Health Care (PHC) system, with units being selected in each municipality using a systematic sampling methodology (71).

Six studies in 957 patients were conducted with the Dutch version (1, 29, 41–44). Five studies of those reported total scores in 785 patients. Thirteen studies in 1,920 patients were conducted with the English version: eight in the USA (6, 22, 46–49, 52, 53); one in the UK; (45) in Canada, the USA and the UK (32); in Nigeria (7), Saudi Arabia (50), and India (51). Ten studies of those reported total scores in 1,490 patients. Seven studies were conducted in 3,677 patients with the Spanish version; six studies used the version psychometrically tested by Martín-Fernández et al. (2,010) (8) and were conducted in Spain (8, 9, 37) and in Mexico (23, 54, 55). One study (56) used the Mexican Spanish translation (10). Five studies of those reported total scores in 2,788 patients. Three Italian studies in 545 patients used the Italian version (57–59). The Romanian version was used in a study conducted in Romania in 55 children and adolescents (60). Eight studies encompassing 2,542 patients used the German version. Four of them reported on the same German representative general population sample rating their primary care doctor (11, 61–63); three more studies were conducted in other German samples (64–66) and one study was conducted in Austria (67). Five studies of those reported total scores in 2,542 patients. One study on the Hebrew version was performed in 93 patients in Israel (33). Two studies with the Arabic version were performed in 1,053 patients in Saudi Arabia (68) and Oman (69). The Persian version was conducted in three studies in 683 patients in Iran (16, 76, 77). Two studies reporting on the Bangla version in 264 patients were performed in Bangladesh (32, 70). The Thai version was used in four studies conducted in 1,205 patients in Thailand (14, 24, 62, 72). Two studies of those reported total scores in 423 patients. The Malay version was used in one study conducted in 906 patients in Malaysia (13). And the Maldivian version was used in the Maldives in 172 patients (9). The Chinese version was used in seven studies in 4,555 patients: five conducted in China (14, 15, 25, 26, 75), one in the Inner Mongolia Autonomous Region of People's Republic of China (74) and one in Taiwan (73). The Afaan Oroma version was used in one study in 372 patients in Ethiopia (30). Two studies on the Portuguese version were conducted in 6,921 patients in Brazil (15, 71).

For 23,343 patients from 50 publications describing 63 samples in 14 languages, the overall weighted mean for the total score on the PDRQ-9 was M_weighted = 34.21, SD__pooled = 6.70. A forest plot based on total scores involving those 63 samples is shown in Figure 2.

As can be seen in Figure 2, the vertical line in the middle is the weighted total mean of all studies. The N per language is reported based on studies reporting an M and SD. The vertical striped lines on the left and on the right represent the SD limits of the weighted total means. The plot is arranged per country and the first dot per country represents the total for that country. As can be seen in the figure, the weighted total mean for the Dutch, English, Hebrew and Chinese versions correspond with the weighted total mean of all studies. The Spanish, Italian, Bangla and Thai version have a higher weighted total mean score. The other countries have a lower weighted total mean score.

Many studies found that the PDRQ-9 presented discrimination problems in the upper part of the scale (8, 9, 11, 23, 38, 56, 61–65, 72). They found a clear left-skewed distribution, indicating that most values were concentrated on the right of the mean. This reveals a strong tendency to use the highest response options of the PDRQ-9 items and rate one's doctor positively. This led to skewed item scores and a skewed instrument's sum score. The latter is shown in Supplementary Figure A1 in the Supplementary Materials with a histogram for PDRQ-9 total scores of the German sample (11). The left side of Supplementary Figure A1 shows the distribution of PDRQ-9 total scores. Based on Rankit normalisation we converted the PDRQ-9 total score to normalised T-scores, whose distribution is shown on the right side in Supplementary Figure A1.

The relation between raw scores on the PDRQ-9 and T-scores under the normal distribution is shown in Figure 3.

The resulting PDRQ-9T-score metric is especially apt at distinguishing severity levels among the mildly to severely challenged PDRs in the lower realm.

19 Studies provided mean item scores of 14,486 patients (1, 7–9, 11, 12, 15, 16, 49–51, 54, 68, 70, 71, 73, 75–77) and 6 studies provided frequencies of response item options (1, 11, 50, 51, 70, 71). An overview of weighted mean item scores and SD__pooled is shown in Table 4.

The table shows the skewed representation of the PDRQ-9 item scores. Higher response options (“Appropriate”, “Mostly”, and “Totally”) were more frequently chosen than lower response options (“Not at all” and “Somewhat”). Furthermore, Table 4 shows data on the frequency of endorsed response options for a total of 14,697 patients as provided by twenty-one studies. (1, 7–9, 11, 12, 15, 16, 41, 49–51, 54, 68, 70–73, 75–77).

The weighted mean scores and pooled SD's for items are shown in Supplementary Table A1.

Several studies suggested single cut-off values for a good PDR: 31 in Brazil (13, 71); >33.57 in Mexico (23); >34.5 in Ethiopia (30); and ≥39.69 in Spain (8), which is a much higher value than the others.

Other studies suggested cut-off values for distinct levels of severity, defining good and low PDR scores as 29 and ≤22 in Mexico (55); as ≥43.4 and ≤30.4 in the USA (53); as ≥36; 18–35 (moderate PDR); and ≤17 in Saudi Arabia (50) and Thailand (38, 72); as >36, 22–35 (moderate PDR) and ≤21 in Iran (77); ≥33, 21–32 (moderate) and ≤20 in the Maldives (35).

One study, conducted in Iran, suggested four levels with cut-off values: ≥38 for excellent “communication skill”, 29–38 for good, 19–28 for moderate, and 9–18 for weak “communication skill” (76).

Based on these reported cut-off scores and given the ceiling effect in the data of the studies, resulting in skewed sum scores on the instrument, distinguishing levels above average and providing a cut-off score is hard to do. Therefore, we provided norms that are especially useful for distinguishing levels below average or challenged PDRs. To this effect, we transformed raw scores from the German general population sample (11) to normalised T-scores and present these for all possible raw scores in the supplementary materials (Supplementary Table A2). This is a crosswalk table to convert raw PDRQ-9 scores to percentile rank order scores, normalised T-scores, and various levels of a challenged PDR. These T scores can also be established with a conversion formula (T = 0.61 + 1.885*RS-0.03865*RS² + 0.0006151*RS³, a 3rd-grade polynomial), have a mean of M = 50.0 and SD = 10.0, and are equidistant. Scores below 30 (raw score <23) indicate an exceptionally challenged relation, reported by <2% of the general population. The Supplementary Table A2 presents a crosswalk table for PDRQ-9 raw scores to T-scores and Percentile Rank order (PR) scores.

Based on these normalised T-scores we recommend the following cut-off values as shown in Table 5: A cut-off value of T ≤ 44 for a challenged PDR, 45 ≤ T ≤ 56 for an average PDR, and T > 56 for a good PDR.

The studies give a range of cut-off scores for severely challenged PDR ranging from 17.00 to 39.69. The relation between raw scores, percentile ranks, and T-scores is also depicted in Supplementary Figure A2. T-scores are on the x-axis, percentile rank-order scores are on the Y-axis, and raw scores are shown in the figure with a white background on the sigmoid curve. Scores in the green area (RS ≥ 35) represent average to good PDRs, and scores in the red area (RS ≤ 30) indicate challenged PDRs. The figure shows how the raw scores in the white squares correspond to T-scores on the X-axis and percentile scores on the Y-axis. The red area indicates challenged PDR scores.

The scoring instructions and formula for conversion of raw scores to standardised, normalised scores is included in the Dutch and English version of the PDRQ-9 in the Supplementary Materials.

Several studies explored challenged PDRs in the following groups as indicated by lower total scores.

In terms of medically unexplained symptoms, one study in primary care found that MUS patients reported significantly higher levels of loneliness and a lower quality of the PDR than older patients with explained medical symptoms (p = 0.002) (46). When taking into account comorbid depressive and anxiety disorders, only the association between MUS and the quality of the PDR remained statistically significant. The effect size of the association was small, with a beta of 0.179–0.191. p = 0.005–0.009. A Chinese study found that patients with mixed somatic and psychological distress had the worst PDR (M = 36.0; SD = 7.3). In addition, compared with patients with somatic distress alone, those with psychological distress alone had a worse PDR (73). Lower than average scores were found in fibromyalgia patients (32), and in primary care patients with persistent unexplained symptoms in primary care with high rates of comorbid anxiety and depressive disorders (41).

Regarding mental health conditions, average PDRQ-9 scores were lower in a study in the Netherlands in the specialised mental health setting (43). The PDRQ-9 was found to be negatively correlated with the PHQ-9 (r = −0.196 to −0.309) and there were significant differences in PDRQ-9 ratings between general hospital in-patients with or without depressive symptoms (p = 0.019) in a Chinese study. This correlation was independent of age (14). Modest positive correlations were found of the PDRQ-9 with functional (p < .001) and global quality-of-life (p < .001), and negative correlations of the PDRQ-9 with psychological distress (p < .001) in oncology patients receiving treatment in specialist general hospital centres. Also, low PDRQ-9 scores for the PDR with the oncologist were associated with more anxiety (p = .006) and depression (p = .004) (9). In a general population sample, adjusted for age, the mean PDRQ-9 score of participants with a potential depressive disorder was M = 3.66 (SD = .86), and that of participants without a potential depressive disorder was M = 4.12 (SD = .66) (F = 65.8, p.001), a difference with a small effect size (Partial eta2 = .05). Experiencing somatic and depressive symptoms emerged as the most relevant patient-related predictors of the quality of the PDR with the family physician (11).

In terms of factors influencing treatment, in a Dutch study, the PDRQ-9 rating of the quality of the PDR was the only variable among patients' characteristics which was significantly and positively correlated (Pearson rho = −0.49; P < 0.01) with decisional conflict (42). Two other studies corroborated that the only factor associated with decisional conflict was the PDR (46, 47). A good understanding of the patients' illness was linked to a better relationship with their doctor and greater participation in shared decision making which was associated with increased trust and adherence to medication in another study (45). A study in psychiatric patients of Hospitals in Istanbul found a correlation between irregular attendance of the mental health services, and lower PDR (135). A Spanish study found a positive association (r = 0.38); p < 0.001) between consultation time and PDRQ-9 score: for every minute increase in consultation time, the PDRQ-9 score increased in 0.04 units (p = 0.001, CI95%: 0.016–0.063) (56).

PDRQ-9 scores were positively correlated with medication adherence rates for 3 months of treatment (r = 0.52; P = 0.006). Subjects in a US study who perceived a weaker PDR (less than or equal to 36, n = 10) were less adherent over 3 months, with an average adherence rate of 45%, compared with subjects who perceived a stronger PDR (greater than or equal to 37, n = 17), with an average adherence rate of 70% (P = 0.03). Adherence did not vary by age or gender (P = 0.59 and 0.51, respectively) (53). Several other studies found a positive correlation between PDRQ-9 score and medication adherence (35, 69). Regarding online treatments, internet hospital users who received online consultations rather than face to face gave significantly lower scores than regular hospital users (P = .01) (15).

A Spanish study found that men had significantly higher PDRQ-9 scores than women (9). In Saudi Arabia, a female sample with chronic disease had an overall level of the PDR that was less than moderate. The PDR had a significant influence on the patients' self-efficacy in self-managing their chronic diseases (p = 0.047) (68). A study involving trans and non-binary patients found that non-binary participants showed a lower score (M = 30.79; SD = 7.46) than trans women (M = 34.78; SD = 9.17) and trans men (M = 36.10; SD = 9.45), a significant difference in means (F = 3.52, df = 2, p = 0.034) (67). The gender of the physicians was reported in none of these studies.

The number of studies exploring the role of age in the PDR was limited. One study in primary care found no significant correlation of the PDR with age, health, or psychological distress (6); another study found higher age (OR 1.03, 95% CI: 1.02–1.05) was associated with above average PDR (8). However, given the OR of 1.03, this effect can be considered to be clinically irrelevant.

Based on Tables 1–3, the following can be resumed regarding cross-cultural validity of the various language versions of the PDRQ-9.

The original Dutch language version's validity was supported by psychometric analyses, its factor structure reported, and means and SDs reported in five out of six field studies. The English version was produced by forward – backward translation and consultation of native speakers; and psychometrically tested in USA, Canadian, UK and Nigerian studies. The one-factor structure of this English version was confirmed by confirmatory factor analysis performed in two of those studies, one in the USA and one in Nigeria (6, 7). This version was used in 11 field studies, 10 of which reported means and SDs. The original published English version can therefore be considered as extensively cross culturally psychometrically tested.

The German version was developed by forward backward translation and applied in a representative sample from the general population. Internal consistency and construct validity were explored and a CFA performed, which confirmed a one-factor structure. Eight field studies were executed, four of them in the general population sample and four in convenience samples, reporting means and SDs. Furthermore, the German version was at the basis of one of the Chinese language versions that was used in two studies, after an extensive translation and psychometric testing procedure (25, 26).

The Spanish version was developed in Spain, with three studies reporting reliability, two reporting validity and factor analysis. One Mexican Spanish version was also developed and investigated with CFA. Three field studies were performed in Spain and four in Mexico, reporting means and SDs.

The Thai version was developed by forward backward translation and CFA reported. Four field studies were done with three reporting means and SD. The Bengali version was developed by forward backward translation in Bangladesh. Factor analysis confirmed one component. Two field studies reported total means and SDs. The Persian version used forward backward translation, performed factor analysis and three field studies reported means and SDs. The Brazilian Portuguese version performed forward and backward translation and expert consultation, performed factor analysis and two field studies reporting means and SDs. The Arab version was made by forward-backward translation. The study was performed in Egypt and reported Cronbach's alpha but no factor analysis (31). Two field studies reported means and SDs, one in Saudi Arabia and one in Oman (68, 69).

Three Chinese versions have been used where it was unclear how these were translated (73–75). Li et al. reported translation method, Cronbach's alpha and factor structure indices aligning with a one-factor structure (15). Wang et al. developed a Chinese version by an extensive forward backward translation process and found support for a two factor structure with highly correlated factors (14). The authors note that their study is the only one reporting two factors and suggest that this might be related to a less individualistic culture in China compared to other countries, which may make Chinese people more apt to not only consider the actions of their doctor in the appraisal of the PDR, but also the contextual circumstances provided by the hospital. This study is the only one showing a cross-cultural difference in factor structure in all the studies performed with the PDRQ-9 so far, differentiating that Chinese version from the Li study. Seven field studies reported means and SDs: all in China, one in Taiwan and one in Mongolia, two of which were developed in collaboration with a German research group (25, 26). Those authors confirmed in writing that they followed an extensive translation process and that CFA confirmed a one-dimensional factor structure and high reliability, which was however only partly published. The remaining studies (73–75) used translations with unknown factor structure.

The following language versions may need further validation and field studies to be conclusive: The Turkish version was developed by expert involvement and CFA reported; another study also reported EFA. However, no field studies reporting means and SDs were done. The Ethiopian study developed an Afaan Oromo Version by forward backward translation and explored internal consistency, but did not report factor analysis or a field study with total means and SDs (30). The Hebrew version used forward and backward translation and reported construct validity and reliability, and one field study with means and SDs. The Malay and the Maldivian versions do not provide results regarding the factor structure. Three studies report on means and SDs of an Italian version of which development, validity and factor structure is not described; one Romanian study also describes means and SDs but no other information regarding development and validity of the Romanian version was provided.

In summary, this shows extensive cross-cultural validity for the factor structure in 9 language versions, and 7 language versions that are well underway and for which more work may be needed. In that respect, it is relevant to mention that an officially translated Italian version is currently developed in a study funded by the European Union. The picture emanating from this shows good cross cultural validity. In total, studies on the PDRQ-9 encompass four out of five continents with overwhelmingly similar structure and results.

Based on our findings, it seems that one-dimensionality of the PDRQ-9 is cross-culturally supported. One Chinese study does not support that, however, the correlation between the two factors that they find is very high (14). The other Chinese studies either do report a one-factor structure (15, 25, 26), or do not report a factor structure at all (73–75).

Another aspect that warrants attention is the variability in mean scores across studies. These differences may be attributable to variations in study populations (e.g., general population vs. clinical samples), characteristics of health care systems, translation-related issues, and broader cultural factors. Given the present data, it is difficult to disentangle the relative contributions of these factors. Therefore, norming based on representative samples from the general population is recommended.

Total scores of the PDRQ-9 are provided in 50 publications describing studies conducted in 63 samples, 23,343 patients and 24 countries, using psychometrically tested versions of the PDRQ-9. The overall weighted mean for the total score on the PDRQ-9 was Mweighted = 34.21, SD_pooled = 6.70. The weighted mean average varies between countries, with scores higher than average for studies conducted in the UK and USA, the Netherlands, Germany, Spain, the Northern part of Italy, and Thailand. This may reflect good health services in those regions compared to means in Hebrew, Chinese, Persian, and Romanian studies, which corresponded more to the weighted mean of the entire set of studies. Studies with the Arab, Romanian, Brazilian (Portuguese), Mexican and Malay versions yielded lower weighted mean scores. There are several possible explanations for this.

First, although most studies were performed with thoroughly translated and psychometrically tested language versions, the weighted total means scores of the Arab version seem to be afflicted by one study that was performed before the rigorously developed Arab version was published. The mean score reported in that study was substantially lower (68). This may be related to the quality of that translation, or with the population of patients with chronic medical conditions. The same may apply to the Romanian study, as it has an unclear translation process; however, given that the Romanian version clearly shows improvement in PDRQ-9 score after an intervention actually underscores the validity of this language version (60).

Second, it could reflect variations in health care service. For example, the M and SDs in the Brazilian sample are much lower than in other studies using the PDRQ-9. Methodological factors could play a role given that in this rural sample, comprising many illiterate patients from impoverished neighbourhoods, the PDRQ-9 was administered by interview, which might make people lean more toward average answers rather than at the high end of the scale, as in other studies that used the PDRQ-9 as a self-report measure. It is good to know that the interview method of administration allows for the valid appraisal of the PDR in deprived areas, but this difference in assessment may have influenced the completion of the PDR. Furthermore, relevant cultural and socio-economic factors could be that this sample contained many people who were illiterate and from impoverished neighbourhoods. The potential influence of such aspects of the primary healthcare system under study in Brazil on the PDR is captured effectively by the PDRQ-9.

However, it could also indicate PDR differences among patient populations. The findings support the known group validity of the PDRQ-9. Average scores tend to be lower in patients with chronic medical conditions (68) including Parkinsons disease (54). For example, the lower scores for the Arab version and the Mexican version might be partially explained by that.

Medically unexplained symptoms (MUS) and comorbid depressive and anxiety disorders, including fibromyalgia patients, are also found to be associated with lower scores (25, 32, 41, 44).

Dissatisfaction with the PDR in these patient groups is a long-standing issue, with many studies initially reporting on this from the perspective of doctors coining such patients as “worried well”, “difficult”, or “heartsink” patients (137–139). Most interventions in these patients would follow a cognitive behavioural treatment approach (140) which would be effective in patients willing and able to follow this treatment. However, studies show that overall the effect size of this approach tends to be small (141) and CBT provision as primary treatment was met over the years with growing discontent of patient groups, with a more recent development being studies with patient-input leading to redefining MUS as medically not yet explained symptoms (MNYES) (142) and emphasizing the need to study aetiology of syndromes, such as for example in chronic fatigue syndrome (143). Given this, the finding that the PDRQ-9 total score is lower than average in several studies in patients with MUS underscores the validity of this questionnaire and the relevance to explore the patient perspective of the PDR in the context of research. This may be supported by the finding of the lower score in the Malay study conducted in people visiting complementary medicine providers, possibly because of dissatisfaction with their regular healthcare providers (34).

Regarding patients with mental health issues, there seems to be a clear effect on the PDR in case of depression or anxiety, especially for patients who receive treatment from medical specialists in the general hospital, for example oncology patients (9); in specialty mental health patients with schizophrenia in Nigeria (7), and in patients with somatoform disorders in the Netherlands (41).

Our findings show that PDR is considered a relevant factor for patient well-being, compliance with treatment, and healthcare delivery in various countries, healthcare settings, and medical specialisations. The PDRQ-9 is especially apt in identifying a challenged PDR, which is the most relevant domain, making it a valuable tool for improving this critical aspect of care. We recommend reporting total scores rather than item scores and applying the normalised T-scores in future studies of the PDRQ-9 and in clinical practice. It could be beneficial to use the PDRQ-9 to screen for potential issues in the PDR in patient groups struggling with unexplained symptoms or mental health issues. This is more relevant as the PDRQ-9 score is associated with other factors influencing treatment, such as decisional conflict, medication adherence, and patients' self-efficacy in managing their chronic disease. The PDRQ-9 can also discern the quality of the PDR in online treatment settings (15) which indicates potential usefulness in evaluating the PDR as experienced by patients in e-Health interventions with the PDRQ-9.

Gender matters for how the PDRQ-9 is completed: Men rate their PDR significantly higher than women. In addition, non-binary participants have lower scores than trans women and trans men, who score average mean levels. However, so far, studies exploring gender have not taken the gender of doctors into account, which we recommend for future studies. Research should also explore which other doctor characteristics are associated with the PDRQ-9 score.

We presented a standardised scoring approach based on a large representative general population sample. This standardisation produces a normalised score that makes the raw scores that are cluttered due to the ceiling effect interpretable as it normalises the scale to equidistant scores. We present this as an example of how raw scores can be normalised. We suggest to apply this standardisation normalisation method of total scores, and the reported cut-scores, when using the PDRQ-9.

Furthermore, we recommend to perform general population studies to confirm norm scores in countries with mean scores diverting from the mean as displayed in Figure 2. New language versions of the PDRQ-9 should preferably be developed using the forward backwards translation method, and report on psychometric aspects including the factor structure.

Limitations of this study are that there was only one representative sample of the general population. Most studies used convenience samples, although they reported on relevant clinical samples. Few studies have reported on gender differences among patients or potential relevant physician factors affecting PDR, such as educational levels, years of experience, gender, and age.

Although the single-factor structure is supported in most primary studies included in this systematic review, a direct cross-cultural comparison or Differential Item Functioning (DIF) analysis of configural, metric, and scalar factorial Invariance among translated versions for different nations or cultures is lacking. Only one of the primary studies assessed configural, metric and scalar DIF within one dataset/country (14), but none have done this across nations or cultures. We recommend inclusion of the PDQR-9 in future cross-country surveys such as the People Living with Chronic Conditions study (PaRIS) (144, 145), the Commonwealth Fund's International Health Policy survey (146), or other international initiatives, as this would secure representative data in many countries, and provide a basis for further research into normative data and generalizable results.

Another limitation concerned the utilization of the Risk of Bias tool. Assessing cross-sectional studies involves unique Risk of Bias considerations, and a recent review found that none of the currently available tools comprehensively address all potential biases pertinent to cross-sectional studies (27). Following the recommendations from that review, we developed a tool specifically for this systematic review. Biases that we should address were representativeness of the sampling frame, non-sampling error from nonresponse or coverage, and whether the outcome, disease or characteristic under study was measured validly and reliably, which in this case was operationalised by the level of validation of the respective translated versions of the PDRQ-9 in the selected studies. This allowed us to establish that Risk of Bias was low in most studies. However, risk of bias was moderate in studies with the Italian, Thai and some Arab versions, which limits the generalisability of the results for those language versions. Nevertheless, all studies show similar results.

Furthermore, there is a lack of uniformity in how the PDRQ-9 is scored: most researchers calculate the recommended sum score, but some calculate the mean score of the nine items. We recommend reporting the sum score in the future to avoid this inconsistency. Finally, the PDRQ-9 has a substantial ceiling effect, translating into a diminished capacity to discriminate among patients who score high on the PDRQ-9. Cut-off scores, therefore, can best be made for lower scores reflecting challenged PDRs and based on normalised T-scores. We recommend applying the normalised T-scores in clinical use and future studies of the PDRQ-9.

The strengths of this study are the large number of studies, the large total sample, and the diversity of the samples regarding cultural background, with four out of five continents of the world represented. The proposed normalisation of scores is innovative, and the use of normalised sum scores enhances the interpretability of PDRQ-9 sum scores even in case of ceiling effects.