Search yield

The systematic search yielded 5991 unique publications that were included for initial abstract screening (Fig. 1). Of those, 272 were included for full-text review and full texts were retrieved for all of these studies. Of the 245 studies excluded following full-text review, 55% were primarily excluded because of a lack of quantitative or qualitative data on delays in the appraisal, help-seeking, diagnosis or treatment intervals. Other studies were excluded because the study sample had an age at illness onset or study participation that was older (20%) or younger (7%) than our target range of 13–24 years; 13% of studies did not report data on a study population or a subgroup of which at least 80% were diagnosed with a bipolar spectrum disorder. The other 5% of studies were excluded because they were not primary research, not journal articles, not published in English or reported data on delays but featured no exploration of factors that could be associated with these delays.

The final yield of studies that met all inclusion criteria was 27 (detailed in Supplementary Table 2). Of these 27 studies, 12 had a primarily cross-sectional design (including chart review and surveys), 7 were cohort studies, 4 were population database studies, 3 were derived from administrative data and 1 was a qualitative study; some studies combined different research designs (detailed in Supplementary Table 2). In our initial search yield, the number of publications dramatically increased year over year in the time period that was included in our search. However, from the studies that met all inclusion criteria, there was an unexpected peak in the 2005–2009 period, after which there were fewer publications that met our criteria (Fig. 2). Of the included studies, three studies were multinational, and overall, ten studies derived data from USA samples, eight from European samples, three from Oceania samples, two each from Canadian and Indian samples, and one each from Egypt and Oman.

Fig. 2

Included studies sorted by publication year.

A narrative summary of findings from the included 27 studies follows below, identifying factors that were or were not found to be associated with delays in the pathway to treatment. Factors that were significantly associated with contributing delays in the pathway to treatment are also summarised in Table 1. Of the 27 included studies, 13 publications featured data relevant to the diagnostic interval, 12 to the pre-treatment interval, 5 to the help-seeking interval and only 2 featured data relevant to the appraisal interval; in terms of factor categories investigated, 19 of the 27 studies reported on patient-related factors; 17 reported on disease-related factors and 11 reported on systemic factors (Supplementary Table 3).

Table 1

Summary of factors identified in included studies that were associated with delays in the pathways to treatment according to individual and related factors (including patient, family and socioeconomic factors), disease and related factors (including manifestations of bipolar disorder and comorbidities), and systemic factors (including clinician and health system factors)

For non-significant trends or pertinent factors that were found to be not associated with delays, see main text.

a. Factor was inconsistently associated with delays.

Appraisal interval

Only two qualitative studies contained findings specific to the appraisal interval. The first reported that some delay of appraisal could result from parental misattribution of early symptoms of bipolar disorder to reactions to childhood adversity, family conflict, grief and trauma. ^{Reference Crowe, Inder, Joyce, Luty, Moor and Carter28} Similarly, early symptoms of bipolar disorder were also misattributed to comorbid substance use disorder (SUD). ^{Reference Crowe, Inder, Joyce, Luty, Moor and Carter28} Parental self-blame for their child’s symptoms also emerged as a common experience in this study. A qualitative study from India found that in comparison to major depressive disorder (MDD), bipolar disorder was more likely to be attributed to supernatural punishment by caregivers. ^{Reference Singhai, Aneja, Swami and Gehlawat29}

Help-seeking interval

Five studies were pertinent to the help-seeking interval, of which five included data on patient factors, two on disease factors and one on systemic factors.

Patient factors identified as contributing to delays in the help-seeking interval included male gender and younger age at onset. However, one study demonstrated a complex association with occupational status and one study found no association with indigenous status. A study of Omani students diagnosed with bipolar disorder identified male gender as having a reduced likelihood of contact with care services. ^{Reference Riyami, Adawi, Kharusi, Morsi and Jaju30} This study also reported that 14- to 16-year olds were less likely to have had contact with care services than those over the age of 17 years. In a cross-sectional study of Finnish youth, having a first mood episode of any polarity before age 18 years was associated with a greater time to first contact with psychiatric care. ^{Reference Suominen, Mantere, Valtonen, Arvilommi, Leppamaki and Paunio31} This was true for both time from first symptoms (11.9 v. 7.2 years) and first episode (10.2 v. 5.5 years). Similarly, in a Canadian population study, 15- to 18-year olds diagnosed with bipolar disorder were less likely to report a lifetime contact with mental health services than 19- to 24-year olds (46 v. 60%); a similar pattern was observed in those who reported a comorbid SUD (23% v. 46%; SUD itself was not analysed as a factor in service use). ^{Reference Kozloff, Cheung, Schaffer, Cairney, Dewa and Veldhuizen32} However, this study demonstrated an interaction between age and occupational status: the global age-related trend was similar among full-time employees (lifetime contact with mental health services reported in 48% of 15- to 18-year olds and 82% of 19- to 24-years olds who were full-time employees), a major inversion was observed in full-time students (81% of 15- to 18-year olds and 23% of 19- to 24-year olds who were full-time students). ^{Reference Kozloff, Cheung, Schaffer, Cairney, Dewa and Veldhuizen32} In New Zealand, indigenous status was not associated with differences in contacting mental health services for affective psychosis. ^{Reference Carr, Cunningham, Petrovic-van der Deen, Manuel, Gibb and Porter40}

The qualitative study of caregivers beliefs about bipolar disorder as compared with MDD in India found that because of the frequent attribution of bipolar disorder to supernatural punishment, more time and other resources would be spent on pursuing traditional faith healing practices. ^{Reference Singhai, Aneja, Swami and Gehlawat29} Similarly, a study from Egypt found that 41% of study participants sought advice from traditional healers (defined as religious or spiritual counsellors), of whom 62% did so before contacting psychiatric services. This was assumed to delay appropriate care; however, an important limitation of this study was that this assumed delay was not explicitly measured. ^{Reference Assad, Okasha, Ramy, Goueli, El-Shinnawy and Nasr33} In this study, seeking help from traditional healers was associated with lower educational level but higher global assessment of functioning; the authors’ interpretation was that greater impairment owing to illness would push individuals to seek psychiatric support. ^{Reference Assad, Okasha, Ramy, Goueli, El-Shinnawy and Nasr33} Gender, age and family history were not associated with seeking help from traditional healers. ^{Reference Assad, Okasha, Ramy, Goueli, El-Shinnawy and Nasr33} This was the only included study that investigated help-seeking in relation to disease factors within the appraisal interval. It found that a history of hallucinations was associated with seeking help from traditional healers (as defined above), whereas having psychiatric comorbidities was associated with a negative history for seeking help from traditional healers. There was no association with accessing traditional healer with bipolar disorder type, episode polarity, episode severity, number of episodes or the presence of delusions. ^{Reference Assad, Okasha, Ramy, Goueli, El-Shinnawy and Nasr33} No other studies were found that explored how help-seeking was associated with neither disease factors nor systemic factors.

Diagnostic interval

Thirteen included studies were pertinent to the diagnostic interval, of which ten had data on patient factors, eight had data on disease factors and six had data on systemic factors.

Patient factors identified as contributing to delays in the diagnostic interval included age (although this finding was inconsistent), an absence of family history, and indigenous or ethnic minority status.

Earlier age at illness onset was associated with delayed diagnosis in two studies. ^{Reference Goldberg and Ernst34,Reference Berk, Dodd, Callaly, Berk, Fitzgerald and de Castella35} However, in a study that offered a more detailed breakdown of the trajectory to diagnosis, earlier age at onset was a factor in the help-seeking interval, but it was not associated with the time from first contact to diagnosis, i.e. the diagnostic interval specifically. ^{Reference Suominen, Mantere, Valtonen, Arvilommi, Leppamaki and Paunio31} This degree of chronological resolution was not provided in the other two studies, wherein the role of age at onset in the help-seeking interval may confound interpretation of their results. In contrast to age at onset, younger age at time of initial contact was associated with a lower likelihood of diagnosis on initial contact in a Danish population study. ^{Reference Kessing36} In the USA context, age was not associated with consistency of insurance claims related to a new diagnosis of bipolar disorder (i.e. diagnostic stability). ^{Reference Evans-Lacko, Riley and Dosreis37,Reference Evans-Lacko, dosReis, Kastelic and Riley38}

Gender was not associated with delays in diagnosis ^{Reference Goldberg and Ernst34,Reference Kessing, Vradi and Kragh Andersen51} or consistency of diagnosis. ^{Reference Evans-Lacko, Riley and Dosreis37,Reference Evans-Lacko, dosReis, Kastelic and Riley38} Race was also not associated with delays in diagnosis in a USA study. ^{Reference Goldberg and Ernst34} However, in New Zealand, indigenous status predicted lower rates of bipolar disorder type 1 diagnosis and higher rates of a diagnosis of schizophrenia following initial presentation of affective psychosis. ^{Reference Carr, Cunningham, Petrovic-van der Deen, Manuel, Gibb and Porter40} Although the data could not confirm that this was attributable to missed recognition of bipolar disorder, the authors considered it the most likely explanation given that other studies have shown similar rates of mood disorders in Maori and Non-Maori patients, as well as underrecognition of mood symptoms and bipolar disorder in other racial or ethnic minorities. ^{Reference Carr, Cunningham, Petrovic-van der Deen, Manuel, Gibb and Porter40} One small cross-sectional study found that absence of a family history of bipolar disorder was the primary risk factor for underdiagnosis of bipolar disorder. ^{Reference Chilakamarri, Filkowski and Ghaemi39}

Key disease factors identified as contributing to delays in the diagnostic interval included absence of psychosis and subjective functional impairment. In one cross-sectional study, the average time to diagnosis more than doubled for individuals whose initial mood episodes were non-psychotic and were subjectively not impairing in terms of academic/occupational or social functioning. ^{Reference Goldberg and Ernst34} Whether illness onset was acute or insidious (recalled subjectively by study participants) had no association with delaying diagnosis. ^{Reference Goldberg and Ernst34} However, greater illness severity was associated with improved diagnostic stability as inferred from continuous insurance claims. ^{Reference Evans-Lacko, Riley and Dosreis37,Reference Evans-Lacko, dosReis, Kastelic and Riley38} Similarly, a greater number of psychiatric comorbidities made at the time of the initial bipolar disorder diagnosis was associated with a more continuous pattern of insurance claims for bipolar disorder. ^{Reference Evans-Lacko, dosReis, Kastelic and Riley38} In contrast, a greater number of psychiatric diagnoses before an index diagnosis of bipolar disorder was associated with a lower degree of claims continuity.

The included studies did not provide clear data demonstrating an association between polarity at illness onset and time to diagnosis. However, this association was inferred by the authors of one study that showed a median onset of depressive symptoms and episode occurring at 18 and 21 years, in contrast to a median onset of manic symptoms and episode at 21 and 24 years of age, respectively. ^{Reference Berk, Dodd, Callaly, Berk, Fitzgerald and de Castella35} This study investigated the diagnostic history of participants diagnosed with either bipolar disorder or schizoaffective disorder. Over half were diagnosed with another condition, most commonly depressive disorders, for a mean duration of 7.6 years before the diagnosis was revised. Similar inferences about a depressive polarity at onset delaying recognition and diagnosis of bipolar disorder were made in two Spanish studies. ^{Reference Soutullo, Escamilla-Canales, Wozniak, Gamazo-Garran, Figueroa-Quintana and Biederman43,Reference Ribeiro-Fernandez, Diez-Suarez and Soutullo52} These studies both showed a high prevalence of depressive symptoms preceding diagnosis. They also observed either a significant drop of depressed mood and anhedonia ^{Reference Soutullo, Escamilla-Canales, Wozniak, Gamazo-Garran, Figueroa-Quintana and Biederman43} or an increase of grandiosity ^{Reference Ribeiro-Fernandez, Diez-Suarez and Soutullo52} at the time when the diagnosis was made.

The included studies did not find a clear association between bipolar disorder type and delays in the diagnostic interval. Only one study identified that, compared to patients with bipolar disorder type 1, there was a significantly poorer diagnostic agreement between referring general practitioners and consulting psychiatrists for patients with bipolar disorder type 2. ^{Reference Daigneault, Duclos, Saury, Paquet, Dumont and Beaulieu53} This difference between bipolar disorder types was not significant in the 18- to 25-year-old age group, although there was a non significant trend for better diagnostic agreement in older age groups. ^{Reference Daigneault, Duclos, Saury, Paquet, Dumont and Beaulieu53}

Regarding clinician and systems factors in the diagnostic interval, lack of systematic evaluation and lack of access to psychiatry was identified as contributing factors. One cross-sectional study from 2005 concluded that lack of systematic interviewing or screening for bipolar disorder was a leading factor for missed diagnoses of bipolar disorder in over half of referred youth initially diagnosed with MDD, although this attribution was not specifically measured. ^{Reference Dilsaver and Akiskal41} Similar conclusions were made in another cross-sectional study from 2007 that a systematic diagnostic interview diagnosed bipolar disorder type 2 in just under one out of five youth originally diagnosed with MDD. ^{Reference Bhargava Raman, Sheshadri, Janardhan Reddy, Girimaji, Srinath and Raghunandan42} In a 2011 cross-sectional study of youth diagnosed with bipolar disorder using a systematic diagnostic interview, one out of three were previously diagnosed with MDD and just over one out of three were diagnosed with attention-deficit hyperactivity disorder (ADHD), whereas only one youth with an initial diagnosis of bipolar disorder was revised to MDD. ^{Reference Chilakamarri, Filkowski and Ghaemi39} Of relevance to these findings, suspicion for bipolar disorder as the reason for referral and diagnosis of bipolar disorder by a psychiatrist was found to be poor in one clinic in Montreal, with an interrater κ-value of 0.3, although data on delayed diagnosis was not available for association with referral reason-diagnosis agreement. ^{Reference Daigneault, Duclos, Saury, Paquet, Dumont and Beaulieu53} In a Spanish cross-sectional study, delays in getting to see a psychiatrist was identified as the primary factor for delays in diagnosis. Concordantly, in the USA context, consistent insurance claims related to a new diagnosis of bipolar disorder (i.e. diagnostic stability) were more likely for youth who received an index diagnosis from a mental health professional than other professionals, and among those who received managed care instead of fee-for-service models. ^{Reference Evans-Lacko, dosReis, Kastelic and Riley38}

Pre-treatment interval

Ten included studies were pertinent to the pre-treatment interval, of which seven had data on patient factors, eight had data on disease factors and five had data on systemic factors. In general, we found it challenging to reliably map the pre-treatment interval onto cross-sectional study designs, as ‘treatment delay’ often incorporated elements of the appraisal, help-seeking and diagnostic intervals, unless these stages were parsed out. Where relevant, we note when results cannot be reliably attributed to the pre-treatment interval.

There were some contradictory findings on the role of age in the pre-treatment interval. In one cross-sectional study, earlier age at onset was associated with a greater time to introduction of a mood stabiliser, although the study design would not have allowed for investigation into the time period specific to the pre-treatment interval. ^{Reference Goldberg and Ernst34} Data from the Stanley Foundation Bipolar Treatment Outcome Network (SFBN) found that individuals with symptom onset in adolescence (aged 13–18 years) had a greater delay to first treatment than those who had onset later in life. ^{Reference Post, Altshuler, Kupka, McElroy, Frye and Rowe44} However, this data was also likely not specific to the pre-treatment interval. In contrast, study designs that enabled investigation of time periods more specific to the pre-treatment interval found the opposite. Insurance claims data that evaluated treatment after a new diagnosis revealed that older youth who saw a psychiatrist leading up to index diagnosis of bipolar disorder were less likely to receive guideline concordant care than younger youth. ^{Reference Dusetzina, Gaynes, Weinberger, Farley, Sleath and Hansen45} Similarly, another study of insurance claims data showed that older youth were less likely to ever fill a prescription and to have a continued pattern of prescription filling. ^{Reference Evans-Lacko, Riley and Dosreis37} This was not replicated in a third study of insurance claims, which found that age was not associated with guideline concordant pharmacotherapy or metabolic screening. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} However, this study did find that older age was associated with lower rates of guideline-concordant psychotherapy and monitoring for side-effects and drug levels. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46}

Socioeconomic factors were also identified as playing an important effect on the pre-treatment interval, although findings regarding the role of patient ethnicity or race were mixed. Low socioeconomic status was associated with non-adherence. ^{Reference DelBello, Hanseman, Adler, Fleck and Strakowski48} In a USA cohort study that focused on the impact of race on treatment initiation, initial treatment adherence was lower in Black patients than in White patients (56 v. 72%). ^{Reference Fleck, Hendricks, DelBello and Strakowski47} Black patients were more likely to be prescribed antipsychotic medications (70 v. 24%) and for a longer proportion of follow-up than White patients, even without psychosis present. ^{Reference Fleck, Hendricks, DelBello and Strakowski47} Rates and duration of antipsychotic prescribing were not accounted for by non-adherence rates; only race and psychosis predicted the rate of antipsychotic prescription. Although subjective evaluations of the helpfulness of treatment was generally associated with better treatment adherence, race was not associated with subjective perceptions of medication helpfulness. ^{Reference Fleck, Hendricks, DelBello and Strakowski47} Race was also not associated with antipsychotic dosing or duration of treatment with antipsychotics after remission. Additionally, race was not linked to rates of mood stabiliser prescription in this same study, ^{Reference Fleck, Hendricks, DelBello and Strakowski47} a finding also observed in another cross-sectional study. ^{Reference Goldberg and Ernst34} However, elsewhere, administrative data found that Black, Asian and minority ethnic youth had lower odds (odds ratio 0.7) of receiving guideline-concordant medications than White youth. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} Race and ethnicity was not associated with blood-level monitoring, side-effect monitoring, metabolic screening or adjunct psychotherapy. Youth residing in non-urban areas were more likely to receive indicated metabolic screening, but less likely to receive adequate psychotherapy. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} Density of residence was not associated with pharmacotherapy, blood level monitoring or side-effect monitoring. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} Gender was not associated with receipt of guideline-concordant care. ^{Reference Goldberg and Ernst34,Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46}

Disease factors were inconsistently linked to treatment delay. One cross-sectional study concluded that neither polarity of first mood episode nor bipolar disorder type was associated with differences in time to treatment. ^{Reference Goldberg and Ernst34} The absence of an association between bipolar disorder type and time to treatment was also observed in another cross-sectional study. ^{Reference Suppes, Leverich, Keck, Nolen, Denicoff and Altshuler49} In contrast, a more recent cohort study showed that the first mood episode for bipolar disorder type 2 was more likely to go untreated within the first 2 years of symptom onset. This may be partly attributable to a later onset of hypomania in bipolar disorder type 2 by 3 years relative to the onset of hypomania or mania in bipolar disorder type 1, or could be related to illness severity, as the first mood episode was treated in an in-patient setting in 45% of patients with bipolar disorder type 1 versus 19% of patients with bipolar disorder type 2. ^{Reference Brancati, Nunes, Scott, O’Donovan, Cervantes and Grof50} However, by design, these three studies have poor specificity to the pre-treatment interval. Although cross-sectional data suggested that disability and psychotic features with first mood episodes shortened delay to treatment, ^{Reference Goldberg and Ernst34} insurance claims data did not support the association with psychotic features at diagnosis and receiving guideline-concordant care. ^{Reference Dusetzina, Gaynes, Weinberger, Farley, Sleath and Hansen45}

Comorbidities were often identified as a strong factor in the pre-treatment interval. SUD ^{Reference Fleck, Hendricks, DelBello and Strakowski47,Reference DelBello, Hanseman, Adler, Fleck and Strakowski48} and ADHD ^{Reference Dusetzina, Gaynes, Weinberger, Farley, Sleath and Hansen45,Reference DelBello, Hanseman, Adler, Fleck and Strakowski48} were associated with lower adherence rates. One study contradicted this, noting that comorbid conditions were not associated with receiving appropriate pharmacotherapy, blood level monitoring, side-effect monitoring or metabolic screening, although still identified comorbid SUD (but not ADHD) as lowering the odds of receiving adequate duration of psychotherapy. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} Patients who developed heavy cannabis use after onset of manic symptoms had a greater duration of untreated mania, whereas pre-existing heavy cannabis use was not associated with duration of untreated mania. Similar associations were identified with duration of untreated bipolar disorder, but these did not reach statistical significance. ^{Reference Kvitland, Ringen, Aminoff, Demmo, Hellvin and Lagerberg54}

Regarding systemic factors, receiving care from a psychiatrist increased the likelihood of guideline-recommended pharmacotherapy initiation. ^{Reference Dusetzina, Gaynes, Weinberger, Farley, Sleath and Hansen45,Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} However, receiving care from a psychiatrist was not associated with drug serum level monitoring, side-effect monitoring, metabolic screening or adjunct psychotherapy. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} Primary care visits related to bipolar disorder were associated with higher likelihood of drug serum level monitoring and metabolic screening. However, primary care visits were not associated with appropriate pharmacotherapy, side-effect monitoring or adjunct psychotherapy. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46}

Initial encounter setting, encounter frequency and encounter focus were identified as important factors. In-patient care was inconsistently associated with receiving guideline-recommended pharmacotherapy, with one study finding a doubling of odds for youth who had an in-patient visit, ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} whereas another study found no association with pharmacotherapy and actually reduced odds of receiving adequate psychotherapy. ^{Reference Dusetzina, Gaynes, Weinberger, Farley, Sleath and Hansen45} Having more out-patient visits increased the odds of receiving appropriate pharmacotherapy and psychotherapy. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} A greater number of non-medication-related mental health contacts was associated with better initial treatment adherence. ^{Reference Fleck, Hendricks, DelBello and Strakowski47} Similarly, psychotherapy was associated with more continuous use of a mood stabiliser, although it was also associated with delayed initiation of a mood stabiliser. ^{Reference Evans-Lacko, Riley and Dosreis37} However, an initial delay in starting a mood stabiliser after diagnosis did not independently predict later continuity of mood stabiliser use. ^{Reference Evans-Lacko, Riley and Dosreis37} Otherwise, psychotherapy was not associated with receiving guideline-recommended pharmacotherapy. ^{Reference Dusetzina, Gaynes, Weinberger, Farley, Sleath and Hansen45} A continuous pattern of diagnosis in insurance claims was also associated with greater likelihood of initiating and continuing mood stabiliser treatment. ^{Reference Evans-Lacko, Riley and Dosreis37}

Youth receiving USA Medicaid through disability rather than poverty were less likely to receive adequate psychotherapy. However, Medicaid eligibility was not associated with pharmacotherapy, drug level monitoring, side-effect monitoring or metabolic screening. ^{Reference Fontanella, Hiance-Steelesmith, Gilchrist, Bridge, Weston and Campo46} The SFBN network study indicated that residing in the USA was associated with more than a doubling of delay to treatment than in Europe, regardless of polarity at onset. ^{Reference Post, Altshuler, Kupka, McElroy, Frye and Rowe44} However, the design of this study may not be specific to the pre-treatment interval and may be confounded by an earlier age at onset in the USA, although no statistical interaction between age and country was found. ^{Reference Post, Altshuler, Kupka, McElroy, Frye and Rowe44}