Consecutive patients and selection bias

Documentation of routine clinical work in consecutive patients leads to an observational cohort without further preselection of the recruited patients. It is important to note that without consecutive screening, patients recruited in trials may not be representative of the disease under investigation, with younger age, male gender, and lower risk groups being more prevalent in conventional cardiovascular trials for example [19]. These authors concluded that consecutive screening should be mandatory in all future clinical trials. Trials without such consecutive screening are considered less reliable and valuable [20]. For this purpose, the research group of our rheumatic outpatient clinic planned a prospective observational trial to recruit all consecutive patients after informed and written consent. The trial was approved by the ethical committee of the Medical University [AN 2017-0041 37074.18 453/AM2 (4581a)].

However, selection biases may occur and must be identified and reported, as they influence the outcome of the statistical analyses and the drawn conclusions. In observational trial of a rheumatic outpatient clinic, a possible selection bias may be the co-existing outpatient clinic for dermatology, ophthalmology, and gastroenterology. Other manifestations of systemic disease may lead a patient to successful treatment of the other comorbidity with subsequent improvement of the rheumatic problem, too. For example, this may apply for patients diagnosed with SpA related to psoriatic disease, uveitis, or inflammatory bowel disease colitis, who were successfully treated using biological or targeted synthetic disease-modifying antirheumatic drugs and, therefore, did not show up in the rheumatic out-patient clinic.

Also, some patients may not provide informed and written consent. This percentage should account for an estimated less than 5% of all patients to provide a reliable base for clinical research.

For interventional trials, it is recommended to randomize patients into treatment and control groups, ensuring the comparability of the two groups regarding their characteristics. For identifying patients with certain diagnostic features, the higher the probability, the larger the size of the consecutive cohort. The clinical, laboratory, and imaging data of these sub-cohorts with specific diagnoses can then be compared. In a consecutive cohort, most patients will show the full spectrum of a disease, but some patients will offer only limited disease expression or even be undiagnosed, fulfilling diagnostic criteria only during follow-up of the disease. Otherwise, some patients with rare manifestations may not show up in such a sub-cohort, but single patients with a less frequent or even orphan or rare disease will show up.

Rapid growth of observational cohorts with consecutive patients

Given the circumstances and localization of the clinical events, such an observational cohort of consecutive patients will rapidly grow. For example, in an ongoing rheumatic out-patient trial, a total of 2,811 patients were recruited and subsequently followed by a single investigator on 1,553 working days between September 2017 and September 2023, with a median of 3 visits [interquartile range (IQR) 1–7] per patient until January 9, 2023 [21]. Inclusion criteria were age over 17 years and the capability to understand the purpose of the study and provide informed consent to the study participation. An additional nine patients drew back their consent after written consent, with data used for the trial only until their withdrawal date. Patients’ data from these and prior visits at this institution were available for the research studies cited in this work and led to a data set of about 20,000 physicians’ reports used for this trial. In this cohort, the primary diagnoses are RA, psoriatic arthritis, and axial SpA accounting for 14.4%, 13.4%, and 11.2% of all diagnoses [21]. These sub-cohorts can then be compared with each other easily.

For this trial, data acquisition was manually performed by students for their diploma works, with all the limitations of possibly missing data and mistakes. Therefore, follow-up studies with larger cohort sizes are difficult to achieve and time-consuming. Based on these and other experiences, such studies are not favoured by clinical researchers, and in the future, with powerful digital health technologies, many analyses can be anticipated with larger cohorts as comparisons during follow-up become easier. Thus, new research and quality assessment dimensions can be developed, even with regular repetitions of low costs.

Patient-centered analyses to drive medicine forward

As a first step of recognizing the value of such a consecutive cohort, single patients with a rare condition, disease, or disease course may be considered for possible publication as a case report. Indeed, from this rheumatic out-patient cohort, cases with significant reduction of pain medication after diagnosis of SpA [22] and the need for accessibility in common variable immunodeficiency with granulomatous-lymphocytic interstitial lung disease (GLILD), who was rapidly and successfully treated with monoclonal antibodies against coronavirus disease 2019 (COVID-19), were already reported [23]. Such case reports may be underrecognized in the literature, but case reports may stimulate the authors to recognize important aspects of their work and the readers to draw attention to specific conditions, diseases, or disease courses. Follow-ups of individual cases may be of specific interest to specialists, then summarized of similar cases to case series even years after publication. Supporting such development by rapid identification and better comparison of health data can be an additional challenge for digital health technologies in the future. Case series with similar disease courses or treatments of patients will further increase the medical community’s knowledge.

Sequential analyses with growing cohort size

For research purposes, cohort studies recruit consecutive patients allowing for cross-sectional observations with or without sequential analysis or sequential hypothesis testing, as the sample size is not fixed in advance. Such research may address prognostic, diagnostic, and therapeutic aspects of patient management. Sequential analyses then provide the opportunity to assess changes during follow-up. From the statistical perspective, with repeated analyses, it is essential to note that as more observations are added, the probability of a Type 1 error increases. Therefore, each interim analysis must adjust the alpha level to control the Type 1 error [24].

Experiences from an observational cohort with consecutive patients

The research question to be followed for analyses of cohort data strongly depends on the cohort size. Initially, the focus will be on frequent diseases, and comparisons between patients’ groups from the same cohort but with different diagnoses are avoided. Next, subgroups of inflammatory and non-inflammatory conditions can be formed. Finally, diagnostic sub-groups with the most frequent diseases can be defined.

With larger sub-cohorts, data can be used to compare with national and international data. Such comparisons allow further conclusions and provide critical information for future studies. For example, a specific focus on cardiovascular risk factors in inflammatory joint disease can be performed, comparing data with Spanish and Norwegian data [25], clinical manifestations of Behcet’s disease can be compared with German and Turkish data [26], and comorbidities in RA patients can be compared with data from the international COMORA trial [27, 28]. Thus, data from national and international cohorts allowed both the confirmation of exciting findings and the discussion of divergent results.

The major limitation of such a study is its retrospective approach, with manual search in unstructured data. Such data acquisition is complex, possibly missing data from additional visits to other institutions. Also, the data set can be too small for age and sex adjustments without the option of repeated analyses in a larger cohort during the follow-up of the ongoing observational trial. In case of too small sample sizes, sequential analyses could even allow a conclusion with improved power with a larger cohort size later [24]. Lower financial and/or human costs will apply when using digital health techniques for these purposes.

Introducing young clinicians into clinical research

The strength of cohort studies is undoubtedly the focus on clinical topics by young researchers who otherwise lack research experience. For example, though differences did not reach a significant level the prevalence of cardiovascular disease was calculated as quite high with 8.7% in SpA, 12.8% in psoriatic arthritis, and 18.7% in RA patients which has to be considered when visiting patients [25]. Behcet’s patients of Turkish origin from the non-endemic Innsbruck sub-cohort present differently compared to those living in Turkey [26]. Polypharmacy defined by ≥ 5 drugs was observed by 33.7% of patients with RA in the Middle-European cohort, compared to 61.6% in the literature [29]; and 88.8% of all Middle-European RA patients were in remission or had low disease activity, with hypertension (38.8%) and osteoporosis (30.0%) as the most frequent comorbidities [27]. The potential need for genetic testing is immense, with 57.3% of patients having the potential to benefit from genetic testing according to their diagnosis and treatment and 53.3% of patients with actually performed genetic testing for diagnostic, prognostic, or pharmacogenetic purposes [21]. The list of such topics can be easily extended, and larger cohort sizes will undoubtedly allow more valid conclusions. With the support of new technologies, young clinicians will learn to approach a clinical topic in their cohort much more easily and compare the results with national and international data.

Use of disease-specific QS

The value of quality assessment is undoubtedly higher when considering disease-specific aspects of a diagnosis. A years-long diagnostic delay from the 1st symptom of inflammatory back pain until diagnosis of SpA may be acceptable compared to a mean international delay of 6.7 years [30], but is unacceptable for patients with RA. In RA, patients have been shown to develop radiographic damage within three months of delayed diagnosis [31].

Only a few disease-specific, international recommendations for improving the quality of health and care services are available. For axial SpA, the Assessment of SpondyloArthritis International Society (ASAS) provided such QS [32]. These recommendations focus on the main aspects of quality improvement: referral times to a rheumatologist, time until a first visit by a rheumatologist, time until diagnosis, monitoring and treatment characteristics, and patient education. Unfortunately, some of the recommended QS are unavailable in the routine clinical assessment, like the time until the general practitioner’s referral to a rheumatologist, the time from referral to the first rheumatologic visit, or the use of the Ankylosing Spondylitis Disease Activity Score (ASDAS) score for monitoring disease activity. Instead, in some institutions, the Bath Ankylosing Spondylitis Disease Activity Score (BASDAI) is still used, which allows rapid decision-making without waiting for laboratory data. Also, physicians’ information about the option of additional medications may not be documented in the charts of all SpA patients with disease activity despite the use of non-steroidal antirheumatic drugs. Accordingly, documentation habits should be adapted to the needs of international recommendations for quality assessments or at least discussed in the international community. Finally published data on all these aspects assessed in rheumatic centers are rare to allow a direct comparison.

Long-term follow-ups of QS over time

Of note, the ASAS report considers the improvement rather than the pure assessment of the quality of health and care services as the primary goal of developing QS. This fact is relevant, especially concerning developing QS over time. Indeed, during the last two decades it appeared in observational cohort, the time from 1st visit to diagnosis in patients with SpA improved, as did the diagnostic delay in patients with RA (Figure 2).

Display full size

Figure 2. 

Possible changes of QS during the past two decades. (A) An observational cohort of consecutive patients with improving times from the first visit until the diagnosis of SpA; (B) an observational cohort of consecutive patients with improving times from the first visit until the diagnosis of RA [25]

General recommendations to be considered

Over the last few years, a growing list of critical international recommendations has been developed, which will further improve qualified clinical practice. Examples of such recommendations are listed in Table 1.

Implementation of all these different recommendations into clinical practice will undoubtedly be a challenge for clinicians and all healthcare workers involved. In parallel, patients will have to be educated and trained on how to use new digital technologies so that they know about their disease, comorbidities, and possible side effects of their medications on a regular, updated basis.