Assessment of patient information needs: A systematic review of measures

Background Providing patient information is a central aspect of patient-centered care. Fulfilling personal information needs has positive effects on several health-related outcomes. Measurement instruments help to identify individual information needs in an effective way. The present study gives an overview of existing information needs measures and further evaluates the quality of their psychometric properties and their psychometric studies. Methods We conducted a systematic search on psychometric studies of measures that assess information needs in PubMed and Embase. Furthermore, we carried out a secondary search with reference and citation tracking of the included articles. Title, abstracts and full texts were screened by two independent reviewers for eligibility. We extracted data on content of the measures, validation samples and psychometric properties. In addition we rated the methodological quality with the COSMIN checklist and the quality of psychometric properties with the criteria of Terwee and colleagues. Results 24 studies on 21 measures were included. Most instruments assessed information needs of patients with cancer or cardiac diseases. The majority of the instruments were in English language and from western countries. Most studies included information on internal consistency and content validity. The ratings showed mixed results with clear deficiencies in the methodological quality of most studies. Discussion This is the first systematic review that summarized the existing evidence on measures on patient information needs using two instruments for a systematic quality assessment. The results show a need for more psychometric studies on existing measures. In addition, reporting on psychometric studies needs to be improved to be able to evaluate the reliability of the psychometric properties. Furthermore, we were not able to identify any measures on information needs for some frequent chronic diseases. Other methods to elicit information needs (e.g. open-ended interviews, question prompt sheets) could be considered as alternatives if sound measures are missing.


Methods
We conducted a systematic search on psychometric studies of measures that assess information needs in PubMed and Embase. Furthermore, we carried out a secondary search with reference and citation tracking of the included articles. Title, abstracts and full texts were screened by two independent reviewers for eligibility. We extracted data on content of the measures, validation samples and psychometric properties. In addition we rated the methodological quality with the COSMIN checklist and the quality of psychometric properties with the criteria of Terwee and colleagues.

Results
24 studies on 21 measures were included. Most instruments assessed information needs of patients with cancer or cardiac diseases. The majority of the instruments were in English language and from western countries. Most studies included information on internal consistency and content validity. The ratings showed mixed results with clear deficiencies in the methodological quality of most studies.

Discussion
This is the first systematic review that summarized the existing evidence on measures on patient information needs using two instruments for a systematic quality assessment. The results show a need for more psychometric studies on existing measures. In addition, reporting on psychometric studies needs to be improved to be able to evaluate the reliability PLOS  Introduction Patient information has been found to be one of the most important dimensions of patientcentered care [1,2]. Meeting information needs increases treatment satisfaction [3] and aids informed decision making [4]. Furthermore, the provision of disease related information is an important determinant for patient-reported health-related outcomes (e.g. treatment adherence, emotional and psychological health/wellbeing, quality of life) [5][6][7]. Studies show that patients often do not receive satisfactory information from their healthcare provider (HCP). [5,8]. Reasons for these findings are multiple. On the one hand, HCPs tend to underestimate patients' information needs and may as well overestimate the amount of information given [5]. Misunderstandings between patient and clinician can lead to missing or false information and/or different appraisal or perception of the importance of information [5,9]. In addition, some HCPs lack the skills to assess the patients' information needs and to provide tailored information. Especially when facing a life-threatening disease, giving the right amount of information to the patient has been found challenging for the HCP as HCPs often assume that too much information might harm [5]. On the other hand, patients themselves have been found insecure or reluctant to voice their personal information needs [5,10].
Nonetheless, providing all potentially relevant information can be problematic too. Too much information can lead to a cognitive overload for the patients. With this overload patients tend to forget relevant information or feel stressed. Not all patients wish for all information [11] and the exact information needs can vary individually [8,12]. Thus, it is important to assess individual needs and tailor information accordingly [10]. Health communication tailored to individual needs enhances the information processed by the patient and supports motivation and behavior changes [13].
The very first step to satisfy individual information needs is to elicit and understand these [14]. Standardized measurement instruments aid to uncover those needs and allow to give relevant information to the patient. For this reason, psychometrically sound measures are needed to assess individual information needs.
Therefore we aimed to identify studies assessing psychometric properties of measures capturing patient information needs, to evaluate the methodological quality of these studies and to assess the quality of the psychometric properties of the identified measures. Overall, this is the first review which provides a comprehensive overview of existing information needs measures and their quality.

Methods
The systematic review was registered in the PROSPERO International Prospective Register of Systematic Reviews (registration code: CRD42014012867). The PRISMA Checklist can be found in S1 Appendix.

Search strategy
We performed an electronic literature search in the databases PubMed and EMBASE, including all articles from their inception to July 16, 2014 and ran an update with the same search strategy on May 31, 2016. The search strategy consisted of a combination of different terms and keywords from the following four domains: (i) construct, (ii) context, (iii) measurement, and (iv) psychometrics (see Table 1). The same search strategy had been used for both databases. No limitations had been applied. Our primary electronic search strategy was complemented by a secondary search, consisting of reference and citation tracking of included articles.

Eligibility criteria
We retrieved peer-reviewed publications, published in English or German. We included studies, which tested psychometric properties of measures that assess the construct patient information needs. Our definition was based on Timmins [8], who defined information needs as "personal expressed needs of the client [. . .] for specific condition-related information. Information needs are therefore expressed needs, rather than normative (defined by the professional)." For the purpose of this review we included measures on expressed needs for treatment-related information as well. We focused on adult patients only and excluded measures that assessed information needs as a subscale of a broader measure. The applied inclusion and exclusion criteria are displayed in Table 2.

Study selection
To facilitate the study selection process, we imported the search results into the reference management software EndNote and removed duplicates. First, two reviewers (WF, EC; update: NK, EC) independently conducted a title and abstract screening to exclude clearly irrelevant records. All articles that were found possibly relevant by one researcher were included. Second,

E1
The full text is NOT available (neither via internet nor library).

E2
The language of publication is NOT English or German.

E3
The publication is NOT an article in a peer-reviewed Journal (e.g. dissertation abstract, conference abstract).

E4
The main aim of the study is NOT to test psychometric properties of an instrument.

E5
The measured construct is NOT information needs.

E6
The instrument does NOT measure information needs of an adult patient (e.g. of relatives, physicians, children, computer user etc.).

Data extraction and quality assessments
Data extraction was conducted by using a data extraction sheet, which was already used in other systematic reviews of measures conducted by the research team [15][16][17][18]. The data extraction sheet contained both descriptive data and data to assess the quality of the included studies. The process of assessing the quality consisted of two distinct steps. First, the methodological quality of the studies was assessed by applying the COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) checklist with a 4-point scale [19]. Second, the quality of the psychometric properties of the identified rating scales was assessed using the quality criteria developed by Terwee et al. [20]. Both rating instruments are described below. To reduce any bias that may occur through single reviewer assessment, two randomly selected studies were extracted and assessed independently by two (EC and IS). The remaining studies were extracted and assessed by one reviewer only (EC).
Assessment of methodological quality. The COSMIN criteria were developed in an international Delphi study, which aimed at developing consensus on definitions and assessments of measurement properties [21]. Nine boxes of the COSMIN checklist [19] refer to methodological standards for studies on measurement properties: internal consistency, reliability, measurement error, content validity, structural validity, hypotheses testing, cross-cultural validity, criterion validity, responsiveness. Each of the boxes consists of several items on design requirements and statistical analyses. The items can be scored on a 4-point rating scale as poor (0), fair (+), good (++) or excellent (+++). To determine the overall score of each box, the lowest score of any item within the box has to be taken (so called "worst score counts" method). For studies based on item response theory (IRT) four items on IRT have to be scored first and the "worst score counts" method has to be applied on each box in combination with those four items.
Assessment of psychometric quality. For the evaluation of the quality of psychometric properties of identified measures we applied the criteria developed by Terwee et al. [20]. These criteria refer to content validity, internal consistency, criterion validity, construct validity, reproducibility (agreement and reliability), responsiveness, floor and ceiling effects, and interpretability. Each of these psychometric properties can be rated by one item as either positive (+), intermediate (?), negative (-), or no information available (0).
Data analysis and synthesis of results. Main characteristics of the included studies and measurements as well as the assessment of the methodological quality and the psychometric quality were combined in a narrative summary. An overview of the results of the quality ratings is also displayed in two tables.

Literature search and study selection
The first original electronic search yielded 6711 records and the secondary searches identified another 55 records. After removal of duplicates, 4889 records remained. Through title and abstract screening 4842 records were excluded and the remaining 47 full-texts were assessed for eligibility. A total of 29 full-texts were excluded according to the predefined criteria (see Table 2) and the remaining 18 full articles were included in the study. In the update of the electronic search 1690 references were found and after duplicate removal 1208 title and abstracts were screened. One article was included from the primary search, and the secondary search added another five articles.
This resulted in 24 included studies overall from the original search and the update. The flow diagram of study selection is displayed in Fig 1. An overview on the reasons for exclusion of full texts is given in the S2 Appendix.
The 24 included studies report on the psychometric properties for 21 distinct measures. Three measures have been psychometrically tested in two studies each. First, this was the Information Needs in Cardiac Rehabilitation measure [36,37]. Second, was a measure on informational needs of men with prostate cancer by Templeton et al. [33] with further testing by O'Connor et al. [31]. Last is the Educational Needs Assessment Tool (ENAT) which had been tested in seven European countries [30] and had been partly reported for the Dutch version in an earlier publication [28]. While twelve measures focus on cancer or cancer treatment information [24-26, 29, 32-34, 38-42], four assess patient information needs related to heart disease [22,35,36,43] and one instrument each on information needs regarding chronic  kidney disease [23], arthritis [27], rheumatic diseases [30], HIV [44] and at discharge from hospital in general [9]. All measures are multi-dimensional and range from nine items [26] to 55 items [36]. Descriptive data of the included measures are shown in Table 4.

Methodological quality of included studies
The assessment of the methodological quality of the included studies by applying the COSMIN checklist is displayed in Table 5. Included studies reported a mean of 2.8 out of the nine COS-MIN criteria. For two studies only content validity could be rated as their measurement used a differential scale (Thurstone scale) [23,26]. Only one study used a measure based on IRT [30] and scored "good" on the four IRT-Items.

Quality of psychometric properties
The assessment of the psychometric quality of the identified measures by applying the Terwee criteria [20] is displayed in Table 6. Most included studies reported three or four out of the nine criteria. Again, two studies were only rated for content validity due to their differential scale [23,26]. In addition, the study by Ndosi et al. [30] had not been rated as the Terwee criteria do not apply to IRT.
All in all, the psychometric properties of the identified measures were mainly rated as "intermediate" with some "positive" ratings. The most commonly assessed criteria was internal consistency in 20 studies with two "positive" results [29,44] and the rest being rated "intermediate". This was followed by content validity which had been assessed in 17 studies with six "positive" [9,23,25,26,40,41] and one "negative" [38] rating while the remaining studies received "intermediate" results.

Discussion
The aim of this review was to systematically identify and assess psychometric studies on measures of patient information needs, to assess the methodological quality of these studies and to investigate the quality of the psychometric properties of the identified questionnaires. To our knowledge, this is the first systematic review that systematically appraises and summarizes the existing evidence on measures on patient information needs. Our systematic literature search revealed 24 studies assessing 21 measures on patient information needs.
The methodological quality of the included studies with regard to quality of design, methods and reporting was predominantly rated from "poor" to "fair". Reliability was evaluated in only seven measures, and if reliability has been assessed, the quality was judged as "fair" at best. The poor reliability assessments may be partly due to the "worst case counts" as a conservative method of the COSMIN checklist, i.e. the lowest score of any item in a box is considered the overall score of that box. Consequently, if a study is of good methodological quality but does not provide this information in the study (rated "fair"), the highest score in that box can only be "fair". This emphasizes the importance of further improvements of the reporting of studies on measurement properties in the literature. Although reported more often, the evidence for content validity was weak and rated as "poor" in eleven measures. Only the measure used in the study of Degner et al. [26] was rated as "good". Moreover, the evidence for the structural validity of the included measures was also weak. In 15 measures, structural validity was not analyzed, the remaining 6 tools were rated "fair" at best. Another weakness was the lack of reported hypotheses testing, criterion validity and whether the tools are sensitive to change. Although there was only limited information on cross-cultural validity (six measures) the ratings predominantly ranged between "fair" and "good". Remarkably, only one study scored "excellent" on at least one psychometric property when using the COSMIN-checklist.
Overall, the rated standards for design requirements and preferred statistical methods (mean of 2.8) showed mostly "poor" quality. The best results were shown for the measure of Ndosi et al. [30] for rheumatic diseases, which is based on the item response theory and had been tested in a large population in seven European countries. A second study with satisfactory quality was the one conducted by Ghisi et al. [37] on an instrument for the use in cardiac rehabilitation.
Similar results were shown when rating the quality of psychometric properties by applying the Terwee criteria. For many of the psychometric properties, there was only limited information available. Exceptions could be seen for the criteria "content validity", on which ratings mostly ranged from "intermediate" to "positive", and the criteria "internal consistency" and "interpretability" on which ratings were predominantly "intermediate". However, these limited quality ratings on the Terwee criteria partly reflect the low ratings on the COSMIN-checklist. It is likely that some of the measures received a low rating due to incomplete reporting within these studies. The questionnaire used in the study by Mesters et al. [29] for cancer patients was judged as psychometrically most promising. From our findings, we clearly see a need for more evaluation studies on existing measures, focusing on psychometric properties that have not been tested satisfactorily (e.g. reliability, content validity) or hardly or not tested at all for most measures (e.g. structural validity, responsiveness). Moreover, more rigorous study designs and better reporting, including adherence to the COSMIN-checklist, are needed. The results of this review show that there is also a need for the development of new measures with good psychometric properties as there are no measures for a range of frequent chronic diseases (e.g. diabetes, COPD and asthma, depression, chronic back pain). Until today, there is lack of measures on information needs for many indications. Therefore, in clinical practice it is also important to concentrate on other forms of assessing patient information needs. Timmins [8] suggested the provision of an open-style interview, in which the clinician systematically asks patients and their caregivers  Systematic review of measures of patient information needs Table 6. Quality rating of psychometric properties with Terwee criteria.

First author, Year
Focus Psychometric properties For exact information regarding the definitions of psychometric properties see [45]. about their information needs on the respective disease and its treatment. Although this is a time-consuming procedure, the result might be more comprehensive than other interventions (e.g. checklists). Another option is the training of patients to enable them to ask their questions. Yet a Cochrane Review showed that interventions (e.g. question prompt sheets) used before consultations for helping patients to address their information needs showed limited benefits [5].

Strengths and limitations of the study
Strengths of this review are the following. First, we conducted a detailed and recently updated electronic search. Second, two reviewers independently assessed all abstracts and full texts. Third, we used both the COSMIN checklist and the quality criteria for good psychometric properties developed by Terwee et al. [20] for quality assessment.
Limitations of this review are, first, that we excluded measures that assessed information needs as a subscale of a broader measure for reasons of feasibility. Second as data extraction is very time consuming, not all data had been extracted by two researchers. Yet to ensure that the data extraction sheet is clear and yields reliable results, two studies had been independently extracted by two researchers (EC and IS) in the beginning. Furthermore if there had been any doubt on how to extract certain aspects of a study the first researcher (EC) had consulted the second researcher (IS). Finally, we limited our search to two databases and might have missed relevant articles. However, our search strategy was very sensitive and we also added a secondary search to detect studies that were not found by the electronic search.

Conclusion
This is the first systematic review of studies on measures that assess information needs. Due to a comprehensive evaluation of the methodical quality of the included studies and the psychometric properties of the measures the results of this review can help researchers and clinicians to choose the right measure for their research purpose or clinical consultation. Moreover, this review shows a strong need for further evaluation and testing of current measures in this field, including the application of standards like the COSMIN checklist. There is also a need for the development of new measures, as our results show that only for a few conditions specific measures on information needs are available yet.