Background

From December 2019 onwards, a novel coronavirus (SARS-CoV-2) spread worldwide. In March 2020, the World Health Organization classified it as a pandemic [1, 2]. The resulting disease from infection with SARS-CoV-2, termed COVID-19, commonly presents in adults as a typical infection of the upper respiratory tract. It ranges from mild to moderate fever, cough and fatigue [3, 4]. In severe cases, pneumonia may develop (15% of cases), whilst an estimated 5% of COVID-19 patients suffer from acute respiratory distress syndrome (ARDS), septic shock and/or multiple organ failure [5, 6]. Furthermore, it is anticipated that COVID-19 may have a major impact on physical, cognitive, mental and social health status, even in patients with mild disease presentation [7].

Although the disease is often acute, one in ten patients can continue to be affected for weeks or months [6]. This so-called "long COVID" can result in extreme fatigue, muscle and joint pain, breathlessness, heart palpitations, loss or alteration of the sense of taste and smell, gastrointestinal distress and problems with attention, memory and cognition [6]. In a continued pandemic state, this is likely to contribute significantly to global morbidity and mortality. However, the majority of COVID-19 research has been focused on the pathogenesis of SARS-CoV-2 and therapeutic strategies, such as vaccines or antiviral drugs [8]. Consequently, there is a lack of formal evidence of any long COVID impact on both the health and quality of life of affected individuals. This represents an important evidence gap [8].

Research into COVID-19 has preferentially focussed on symptomology, the acute nature of this illness and involved interventions such as therapeutics and vaccines. However, many patients display persistent symptoms with a continued impact on their quality of life weeks and months after the initial disease. Research on the impact of long COVID on quality of life is scarce and, until now, without a validated disease-specific instrument.

Current tools being developed, such as a clinician facing prognostic communication tool with COVID-19 and critical illnesses [9], appear to be clinician-focused. As such, they fail to address the potential impact of COVID-19 on patients' lives beyond its symptoms. Furthermore, these tools are not sufficient to capture QoL, as they lack the sensitivity to capture the often complex and multi-dimensional nature of QoL and its changes [10]. The same is true of non-disease-specific QoL instruments being used in preference to disease-specific ones due to the lack of neutrality of the indicators used [11,12,13].

Hence, this necessitates the development of a patient-centred long COVID specific quality of life instrument. As with other diseases, for one reason or another, the majority of this disease population lie beyond the reach of a clinical trial. Thus, researching this real-world population continues to be of critical importance to a range of interested parties. This includes regulators, payors and prescribers. Their vehicle of choice is the patient registry or observational cohort clinical study of a prospective and/or retrospective type [14].

Study aim and objectives

The overarching aim of this project is to create a disease-specific (long COVID) quality of life instrument to complement existing tools and ongoing initiatives geared towards improving the quality of life of people affected by COVID-19.

Objectives

Hence, the objectives of this research are to:

  1. 1.

    design a post-acute COVID-19 Quality of Life (PAC-19QoL) instrument for the assessment of individuals with long COVID.

  2. 2.

    Validate the developed PAC-19QoL instrument using healthy volunteers.

  3. 3.

    Implement the developed PAC-19QoL instrument as the core dataset in the associated PAC-19QoLReg patient registry (ClinicalTrials.gov Identifier: NCT04586413). The patient registry is currently recruiting participants with a clinical diagnosis of long COVID irrespective of test confirmation for a 12-month follow-up period.

Methods

Study participants

Recruitment took place in December 2020 from adverts placed on social media sites, informal online support groups and snowballing via recruited subjects promoting the study within their own networks of long COVID sufferers. A total of 15 participants, who had either a diagnostic or antibody test confirmation for SARS-COV-2 and were still suffering from post-acute symptoms of COVID-19, were recruited to the study group. Sixteen healthy volunteers participated in the study as the control group to validate the developed PAC-19QoL instrument.

Sample size

The sample size calculation was informed by the need to both achieve saturation of PAC-19QoL indicators and assess the degree to which each indicator differentiated between the disease and non-disease state. Previous experience with the Jandhyala Method, predicted saturation of unique indicators at a minimum target sample size of 10 with an upper limit of 20. For validation, using a univariate approach, an indicator was deemed to differentiate between the two groups if the prevalence of the indicator was > 50% in one group than the other. A minimum sample size of 15 participants per group was calculated to provide sufficient power (at least 80%) to detect this difference of 50% between the two groups at a 5% level of significance.

Identification of PAC-19QoL quality of life indicators (variables) and summary of the Jandhyala method

Using the Jandhyala Method, the PAC-19QoL indicators (QoLIs), also referred to as variables, were identified [15]. The Jandhyala Method is a novel method for observing proportional group awareness and consensus on responses arising from a list-generating questionnaire. The method has been differentiated from competing consensus generating methodologies [16]. The Jandhyala Method has been validated against Delphi, non-Rand modified Delphi and Rand Appropriateness Method in a systematic literature review and was found to be unique in observing consensus and measuring awareness of subject matter across experts. The Jandhyala Method also improves upon the traditional Delphi-style methodologies, through the introduction of new insights into awareness of subject matter in the expert group. It employs a highly efficient two-round anonymised survey approach without any face-to-face interactions between experts.

The participant consensus is achieved by observing levels of awareness and consensus relating to a list of recommended QoLIs for PAC-19QoL. These are solicited via two anonymised online surveys and calculating an awareness index (AI) and consensus index (CI) for each item, respectively. The responses to the awareness round questionnaire were used to assess knowledge awareness by calculating the frequency of each coded item in relation to the overall most frequently occurring coded item (the Awareness Index). Consequently, the consensus index is measured as the percentage of participants supporting the included item, indicating agree or strongly agree to the included item.

The AI and CI, both continuous variables, were further categorised into 4 Awareness and Consensus scores: Complete Awareness or Consensus (AI or CI = 1.00) – A1 or C1; Awareness/ Consensus + (0.50 < AI or CI) – A2 or C2; Consensus – (0 < AI or CI ≤ 0.50) – A3 or C3; and No consensus (AI or CI = 0) – A4 or C4 [15].

Operationalisation of the Jandhyala method in this study

During the first Awareness Round (1) survey, participants were asked to respond to a series of demographic questions, and two main list-generating questions. Participants were asked to provide a minimum of three and a maximum of ten free-text answers. Please refer to Appendix 1 for the instructions given to the Awareness Round (1) survey participants.

The participants' responses from this Awareness Round (1) were analysed per group. They were then refined into a mutually exclusive list of quality of life indicators by three researchers using a process of content analysis and open coding [17, 18]. The codes were then attributed to the relevant participants' answers by one researcher and were confirmed by a second researcher.

The participants who completed the first round were asked to participate in the second Consensus Round (2) survey. They were asked to rate their level of agreement with the inclusion of the QoLIs arising from the Awareness Round (1) survey, using a five-point Likert scale (Strongly agree, Agree, Neither agree nor disagree, Disagree and Strongly disagree). Quality of Life indicators reaching a consensus level of > 50% (CI > 0.5) were retained in the final list and used to populate the PAC-19QoL.

Variables: operational definition and scale

Following the identification of the quality of life indicators, these were included in the PAC-QoL survey instrument as the variables. Study participants were required to rate the variables (QoLIs) using the Likert Scale. The variables included in the PAC-19QoL survey instrument, and their operational definitions are shown in Table 2.

Statistical analyses

Demographics

Using a Chi-square test, the binary presence or absence of the following discrete characteristics were compared between the control and study groups: sleep apnoea, sleeping difficulty, staying asleep difficulty, allergies, assistance with self-care needs, cancer diagnosis, current smoker, doing own shopping, difficulty falling asleep, former smoker, gender, immunosuppressant drugs, long COVID in the past, long COVID symptoms, major surgeries, mobility issues, non-prescribed or homoeopathic medications, organ transplant, other family history condition, physical activity, pregnancy weeks, prescribed medications.

The following categorical increasing grades of pre-existing conditions were compared between both groups using a Chi-square test: "no versus mild/moderate/severity" (asthma, diabetes, high blood pressure, cholesterol, chronic obstructive pulmonary disease (COPD), cystic fibrosis, stroke). Due to the type of statistical analysis conducted, weight categories were compared "underweight/average" versus "obese", ethnicity categories were compared "white" versus "Asian/other", whilst the category "other", in gender demographics, was not considered during the calculations.

A Mann–Whitney U test (p < 0.05) was performed, comparing the following characteristics: age, weight, height, smoking duration, time since quitting smoking and number of cigarettes smoked per day.

PAC-19QoL QoLIs validation

Using a Mann–Whitney test, statistically significant differences between the mean Likert score for each quality of life indicator (variable) were compared between the responses from study participants (both patients and control groups). A p value < 0.05 indicated a statistically significant finding in the presented analyses. All statistical analysis were conducted in R, version 3.6.3.

PAC-19QoL validation

The PAC-19QoL instrument was validated using a control group with 16 healthy individuals recruited from the networks of the researchers. Participant demographics were recorded for the study patients and control populations, and they were required to complete the same questionnaire as the participants with a confirmed COVID-19 test. In addition, data generated from the control group were analysed using a Mann–Whitney test. This approach of using healthy volunteers and the Mann–Whitney test to validate a disease-specific quality of life instrument is a widely used scientific approach [19,20,21,22].

Public and patient involvement

Through our ongoing work, we have established extensive networks with stakeholder groups and service users with rare diseases (such as XLH). In developing this project, we informally discussed the idea of developing a quality of life measure for people with a lived experience of COVID-19 with people within our extensive network. In total, the project idea was discussed with seven people, and each of these people expressed a need for a patient-centred Quality of Life measure that is easy to use and applicable to every aspect of their life, beyond the disease.

Results

Description of study participants

For the development of the PAC-19QoL, 15 long COVID patients were recruited; four (27%) and 11 (73%) were male and female, respectively. The average age of the study patients and control groups were 40 and 35, respectively (p = 0.308). In order to validate the PAC-19QoL, 16 unaffected participants were recruited. Nine (56%) of these were male, six (38%) were female, and one (6%) preferred not to say.

On assessing all baseline characteristics for heterogeneity, statistically significant differences were found between the two groups in the following demographics: long COVID symptoms (p < 0.001), COVID-19 in the past (p < 0.001), allergies (p < 0.05), asthma (p = 0.007), sleeping difficulty (p = 0.007), prescribed medications (p < 0.001) and dietary habits (p < 0.001). The study and control groups were comparable for all other demographic characteristics (Table 1). For all the analyses run in this study, a p value of 0.001 (Chi Square test) has been established, based on study participants experience of COVID-19 symptoms.

Table 1 Demographic characteristics of both control and study groups in the study
Full size table

Generation of quality of life indicators (variables) for inclusion in the PAC-19QoL instrument

The study included 15 participants with long COVID-19, and saturation of unique quality of life indicators was achieved by participant 9 (Appendix 2). Forty-nine unique indicators were generated during the Awareness Round (1) of the Jandhyala Method and grouped into the following four domains and 19 subdomains. These are:

  1. 1.

    Psychological (Mood, Isolation, Motivation, Anxiety, Cognition, Expression, Mental Exertion),

  2. 2.

    Physical (Exertion, Pain, Travel, Somnolence, Smell/taste, Breathlessness, Fine motor, Libido),

  3. 3.

    Social (Isolation, Relationships, Hobbies), and

  4. 4.

    Work (Ability to work)

Anxiety (Psychological domain) and Exertion (Physical domain) were the two most-populated subdomains, containing 16.3% (8/49) and 14.3% (7/49) of QoLIs, respectively. Of the 49 variables (QoLIs), 8 (16.3%) displayed an AI > 0.50. When the full list was presented to the participants in the Consensus Round (2), 48/49 (98%) achieved a relative degree of prompting. Since five QoLIs failed to reach the cut-off point of CI > 0.50, the remaining 44 QoLIs were included in PAC-19QoL. Validation of the PAC-19QoL instrument with regards to its specificity towards patients with long COVID showed that nine out of the 44 QoLIs failed to demonstrate a statistically significant difference between the patient and control groups (Table 2 and Fig. 1). The full list of 44 QoLIs was then converted to variables to populate the finalised PAC-19QoL instrument.

Table 2 The 44 QoLIs meeting threshold CI > 0.05 (CS:1&2) included in the PAC-19QoL and the Mann–Whitney test of difference in Likert means for each item between the study and control populations
Full size table
Fig. 1
figure 1

PAC-19QoLI Likert scores for control and study groups. †QoLIs with no statistically significant differences between groups

Full size image

Discussion

The PAC-19QoL instrument for the assessment of the quality of life in patients with long COVID, was developed using the Jandhyala method to observe consensus on quality of life indicators solicited in response to the questioning of recruited patients on how long COVID has affected their quality of life. It also provided an insight into the distribution of quality of life indicators in their overall initial awareness among the study group and the final consensus. In this regard, 44/49 (89.98%) indicators were observed to have been prompted from below the awareness threshold to above the consensus threshold and therefore deemed appropriate for inclusion in the PAC-19QoL.

This high rate of prompting may reflect the limited ability of the participants to engage with the initial open-ended question due to the increased mental exertion involved in reviewing the areas of their life affected and then providing indicators. In contrast, the consensus round can be argued as a less intensive exercise with the subject being required to select a level of agreement on a pre-populated list. Understanding the intensely limiting impact long COVID has on cognition and mentation can help inform approaches to engaging with this patient population with information. This impact has been highlighted by other studies conducted among patients affected by COVID-19 [23, 24].

Perhaps unsurprisingly, the most populated subdomains were: Psychological > Anxiety and Physical > Exertion. The latter, along with the Psychological > Mental Exertion subdomain, is consistent with the reporting of post-exertional malaise in response to both physical and mental exertion by other researchers [23]. The former brings into focus the specific concerns long COVID generates in its sufferers on their future. Findings of post-traumatic stress disorder (PTSD) and idiopathic anxiety informs targets for supportive psychiatric, psychological interventions among a portfolio of multidisciplinary management strategies. The 44 quality of life indicators (variables) reaching the consensus threshold for inclusion in the final list of the PAC-19QoL instrument used the basic 5-point Likert scales to assess the severity of each indicator at the time of administering.

Long COVID-19, is a multisystem disease characterised by a range of symptoms, (disease indicators experienced by the patient) and clinical signs (disease indicators observed or elicited by the clinician) aligned to each of these body systems. In this study, some of the included quality of life indicators were observed to align with a body system through a potential aetiological link. For example, QoLI 32 (Table 2, ‘Effect on the ability to converse due to breathlessness), while identified in this study as a physical domain indicator, may have clinical origins in both the respiratory and cardiovascular systems [25]. Similarly, QoLIs 16,17 and 18 on mental exertion (Table 2, Psychological domain) have similar clinical associations with the neurological system. This being said, some included QoL indicators do not readily align to these body systems, most notably those in the social and work domains. This emphasises not only the fundamental differences in conceptualisation of QoL and clinical constructs but also the need to consider them separately, as not all of the clinical symptoms and signs associated with Long COVID-19 will impact sufferers’ quality of life and not all the QoLI’s will clinically manifest themselves [26]. Thus, QoL is a highly relevant but discrete construct that should be measured in Long COVID-19 patients to fully appreciate the impact of the disease with Neutrality.

On completion of its validation, the PAC-19QoL was implemented in the associated PAC-19QoLReg. This patient registry is currently recruiting and will observe a cohort of participants with long COVID, including those with a clinical diagnosis without a confirmed test, over a period of 12 months, with the PAC-19QoL being administering on a monthly basis. It offers the opportunity to track the quality of life of the participants with long COVID beyond simple symptom monitoring, although any relapsing or remitting characteristics will add valuable knowledge to this emerging and debilitation disease.

Future work

The PAC-19QoL instrument as a disease-specific tool is a reliable tool that can be used to effectively measure the impact of long COVID on the quality of life of affected patients. There is need for future work to use the tool in different countries, across cultures, and different languages. This is likely to provide further insight into the validity and reliability of the PAC-19QoL instrument.

Limitations of the study

An initial challenge to the recruitment of subjects to this study was the requirement for a positive test to ensure the indicators generated could be reasonably attributed to long COVID. This requirement generated a degree of frustration in potential participants, as a key concern around their initial management was the strict instructions of not presenting at their primary or secondary care facilities, thereby negating access to any form of testing.

A second potential limitation can be addressed by the fact that a number of QoLIs failed to differentiate between the affected and non-affected individuals. These indicators perhaps infer broader concerns around long COVID not limited to sufferers. Furthermore, these may relate to the general impact of the restrictions imposed to control the spread of the virus, e.g., 'physical isolation from family members' and 'ability to use public transport'. Quality of life indicators such as 'low mood' and 'anger', 'anxiousness about future health of children' and 'future financial situation' are understandable concerns for anyone living through a global pandemic of a novel virus.

A debate over whether the inclusion or exclusion of the quality of life indicators impacts the validity of the PAC-19QoL may ensue. While exclusion will remove any commonality of QoLIs between groups, thus reducing a false-positive rate, removing relevant QoLIs is likely to increase the false-negative rate. Given the overall proportion of these 9/44 (20%) and the damaging effect of not detecting and following up an individual with long COVID, a reasonable justification can be made for retaining them.

Another potential limitation is the number of participants, Although the numbers of participants included in the study were sufficient to test univariate associations, studies with larger number of participants are required to test multivariable associations towards confirming these results.

Finally, there is a potential for bias in the QoLIs included in the PAC-19QoL instrument. This due to patient characteristics, particularly the difference in the educational status of patients, and how this might affect the quality of their responses and contribution to the unique quality of life indicators generated during the Awareness Round (1) of the Jandhyala Method.

Conclusions and implications

It is hoped that the successful development and validation of the PAC-19QoL, a long COVID disease-specific quality of life instrument and its implementation in a dedicated patient registry (PAC-19QoLReg) will complement ongoing research initiatives in monitoring long COVID QoL progression. It is also hoped that the development of the instrument will help to detect responses to therapeutic interventions with greater accuracy, ultimately informing patient care and improving outcomes.